Thermal Adaptation: A Theoretical and Empirical Synthesis

Thermal Adaptation This page intentionally left blank Thermal Adaptation A Theoretical and Empirical Synthesis Mich...

Author: Michael J. Angilletta Jr.

106 downloads 1650 Views 6MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Thermal Adaptation

This page intentionally left blank

Thermal Adaptation A Theoretical and Empirical Synthesis Michael J. Angilletta Jr.

1

3

Great Clarendon Street, Oxford ox2 6dp Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With ofﬁces in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York © Michael J. Angilletta Jr. 2009 The moral rights of the author have been asserted Database right Oxford University Press (maker) First Published 2009 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Data available Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India Printed in Great Britain on acid-free paper by CPI Antony Rowe, Chippenham, Wiltshire ISBN 978–0–19–857087–5 10 9 8 7 6 5 4 3 2 1

978–0–19–857088–2

To Amy, Taylor, and Kyle, for your exceeding tolerance of my wild ideas.

This page intentionally left blank

Preface

The ﬁrst book that I read as a graduate student was Foraging Theory (Stephens, D. W., and J. R. Krebs. 1986, Foraging Theory. Princeton, Princeton University Press). At the time, I was planning to study the evolution of foraging behavior in snakes, and thus wanted to acquire a conceptual background for this work. As fate would have it, my subsequent research did not focus on foraging behavior (or snakes), but the book made a lasting impression on me anyway. The authors laid out a comprehensive and insightful theory, moving from simple models to complex ones, while drawing on relevant empirical data. The book culminated in a rigorous appraisal of the ﬁeld’s accomplishments, driven by the match between theoretical predictions and empirical observations. Had I pursued my interest in foraging biology, the book would surely have become my academic bible. Instead, I soon developed a greater interest in thermal biology, only to discover that no analogous resource existed for this discipline. I don’t want to give you the impression that little was known about thermal biology or that no one had attempted a conceptual synthesis. Neither impression would be accurate. In many respects, thermal biology has progressed tremendously, as one can conclude from reading an early synthesis of this topic (see Cossins, A. R., and K. B. Bowler. 1987, Temperature Biology of Animals. New York, Chapman and Hall). In other respects, though, thermal biology continues to lag behind other disciplines. Let me explain. When it comes to documenting patterns, few disciplines have made more progress than thermal biology. But if we judge progress by our understanding of the processes that generate these patterns, then we have much progress to make. While proximate explanations abound, ultimate explanations remain scarce. This problem stems

from the paucity of quantitative models of thermal adaptation. Thermal biologists have described patterns and documented effects to the point that empirical knowledge has far outstripped theoretical understanding. Most papers in thermal biology focus on empirical descriptions or rely on verbal arguments about adaptation. In comparison, consider the discipline of foraging biology that I referred to above; herein lies a trove of quantitative models and experimental tests about evolutionary hypotheses. The basic ideas about the evolution of foraging behavior were formulated in the 1960s and have been gradually developed to the point that the evolutionary impacts of many factors—from constraints on digestion to risks of predation—have been explored in great detail. Few evolutionary processes in thermal biology have been modeled to the same degree, and no major synthesis of evolutionary thermal biology has emerged. My aim is not to cast blame (for I would certainly share in this blame). Rather, I want to see thermal biology keep pace with other biological disciplines. And to do so, we must develop a quantitative theory of thermal adaptation. Upon closer inspection, a wealth of evolutionary models lie waiting for thermal biologists to adopt as their own. Beginning with Levins (Levins, R. 1968, Evolution in Changing Environments: Some Theoretical Explorations. Princeton, Princeton University Press), biologists have been modeling the evolution of quantitative traits in variable environments for more than four decades. With the growth of this theory, an increasing emphasis has been placed on genetic, developmental, and ecological issues. Many of the models can be applied to understand aspects of thermal adaptation. Although some of these models are mathematically intense, I have distilled from them the essence of what they tell us about adaptation to temperature. For thermal vii

viii

P R E FA C E

biologists, this book contains a synthesis of theory that should guide research in the coming decade. For evolutionary biologists, the book contains an evaluation of current models of evolution in variable environments. Throughout the book, you may notice a taxonomic bias. This bias stems from two factors. First, intentional bias resulted from my conceptual focus. The book revolves around the concept of a thermal performance curve, which describes the relationship between body temperature and organismal performance. Most of the book draws on studies of ectotherms because we can measure their performance curves more readily than we can measure those of endotherms. Students of endotherms will ﬁnd the most relevant theoretical and empirical content in Chapter 4, but other chapters contain relevant theory. Second, some bias inevitably results from the lack of suitable data for certain taxa. In each chapter, I focused on the data most relevant to the theory. Consequently, certain chapters are rich with examples involving animals, plants, or microbes, but no chapter contains an even coverage of all of these groups. I hope this taxonomic bias reﬂects my effort to evaluate the theory more than my ignorance of the literature (although the latter seems inevitable given the exponential growth of knowledge). Above all, I tried to produce the kind of synthesis that I had longed for when I was a graduate student. If I succeeded in this effort, my success was due partly to the biologists who mentored me in my formative years: Justin Congdon, Art Dunham, Henry John-Alder, and Peter Petraitis. My views were also shaped by discussions or collaborations with George Bakken, Lauren Buckley, Ray Huey, Rob James, Joel Kingsolver, Bill Mitchell, Carlos Navas, Amanda Niehaus, Peter Niewiarowski, Warren Porter, Leslie Rissler, Mike Sears, Frank Seebacher, and Robbie Wilson. Many colleagues were kind enough to comment on sections of the book or provide information about their research: David Atkinson, George Bakken, Lauren Buckley,

Brandon Cooper, Lisa Crozier, Marcin Czarnołeski, Wilfried Gabriel, George Gilchrist, Lumír Gvoždík, Gabor Herczeg, Luke Hoekstra, Ray Huey, Jose Martín, Bill Mitchell, Mike Sears, Ben Williams, and Robbie Wilson. Their suggestions and insights contributed greatly to the ﬁnal product. Other colleagues provided original images or permission to adapt ﬁgures: Martin Ackermann, Steve Beaupre, Gabriel Blouin-Demers, Lauren Buckley, Andy Clarke, Lisa Crozier, Michael Dillon, Wilfried Gabriel, Ted Garland, George Gilchrist, Bruce Grant, Brian Helmuth, Ray Huey, Jacob Johansson, Joel Kingsolver, Mark Kirkpatrick, Michael Kopp, Sue Mitchell, Carlos Navas, Mark New, Camille Parmesan, Steve Piper, Mike Sears, Tim Shreeve, Gian-Reto Walther, and Kazunori Yamahira. Finally, the preparation of this book was greatly accelerated by the help of Angela Borchelt and Phaedra Seraphimidi, who obtained copyright permissions for the ﬁgures adapted from other sources. An undertaking of this magnitude requires the support and encouragement of many people. I am greatly indebted to my colleagues at Indiana State University, who picked up the slack for me during my sabbatical. The chair of my department, Charles Amlaner, kept my teaching load in check during the remainder of my writing. The Lilly Foundation and the Ofﬁce of the Provost at Indiana State University provided ﬁnancial support. I owe an equal debt to the students who kept my lab productive despite my reallocation of energy from research to writing: Dee Asbury, Brandon Cooper, Joe Ehrenberger, Matt Schuler, Somayeh Semati, Melissa Storm, and Ben Williams. Most of all, I thank my family—Amy, Taylor, and Kyle. They kindly endured my crazy work schedule and my occasional stress responses. I would not have completed this book without their patience and understanding. Michael J. Angilletta Jr. May 24, 2008 Terre Haute, Indiana, USA

Contents

Preface

vii

1

Evolutionary Thermal Biology 1.1 The challenge of evolutionary thermal biology 1.2 Thermal reaction norms 1.3 The role of theory 1.4 Theoretical approaches to evolutionary thermal biology 1.4.1 Optimality models 1.4.2 Quantitative genetic models 1.4.3 Allelic models 1.4.4 The complementarity of theory 1.5 Empirical tools of the evolutionary thermal biologist 1.5.1 Quantifying selection 1.5.2 Experimental evolution 1.5.3 Comparative analysis 1.6 Conclusions

1 1 3 5 6 6 8 10 13 13 14 14 15 18

2

Thermal Heterogeneity 2.1 Operative environmental temperature 2.2 Global patterns of operative temperature 2.2.1 Latitudinal clines 2.2.2 Altitudinal clines 2.3 Quantifying local variation in operative temperatures 2.3.1 Mathematical models 2.3.2 Physical models 2.3.3 Statistical models 2.4 Conclusions

19 20 23 23 27 29 29 31 32 33

3

Thermal Sensitivity 3.1 Patterns of thermal sensitivity 3.2 Proximate mechanisms and tradeoffs 3.2.1 Thermal effects on enzymes (and other proteins) 3.2.2 Membrane structure 3.2.3 Oxygen limitation 3.2.4 Conclusions from considering proximate mechanisms 3.3 Optimal performance curves 3.3.1 Optimal performance curves: survivorship and related performances 3.3.2 Optimal performance curves: fecundity and related performances 3.3.3 Contrasting the two models

35 35 40 41 43 44 45 47 47 51 53 ix

x

CONTENTS

3.4

Using models to understand natural patterns 3.4.1 Survivorship 3.4.2 Locomotion 3.4.3 Development 3.4.4 Growth 3.4.5 Reproduction 3.4.6 Why do certain patterns differ from predicted ones? 3.5 Have we mischaracterized thermal clines? 3.5.1 Reciprocal transplant experiments 3.5.2 Laboratory selection experiments 3.5.3 Conclusions from reciprocal transplant and laboratory selection experiments 3.6 Have phenotypic constraints been correctly identiﬁed? 3.6.1 A jack of all temperatures can be a master of all 3.6.2 The proximate basis of performance determines tradeoffs 3.7 Do all performances affect ﬁtness? 3.8 Does genetic variation constrain thermal adaptation? 3.8.1 A quantitative genetic model based on multivariate selection theory 3.8.2 A genetic model for survivorship and related performances 3.8.3 A genetic model for fecundity and related performances 3.8.4 Predictions of quantitative genetic models depend on genetic parameters 3.9 Does gene ﬂow constrain thermal adaptation? 3.10 Conclusions

54 56 62 66 66 68 69 70 70 71 74 75 75 75 76 79 81 81 83 84 85 87

4

Thermoregulation 4.1 Quantifying patterns of thermoregulation 4.2 Beneﬁts and costs of thermoregulation 4.2.1 Beneﬁts of thermoregulation 4.2.2 Costs of thermoregulation 4.3 An optimality model of thermoregulation 4.4 Do organisms thermoregulate more precisely when the beneﬁts are greater? 4.5 Nonenergetic beneﬁts of thermoregulation 4.5.1 Thermoregulation during infection 4.5.2 Thermoregulation during pregnancy 4.6 Do organisms thermoregulate less precisely when the costs are greater? 4.7 Nonenergetic costs of thermoregulation 4.7.1 Aggressive interactions with competitors 4.7.2 Risk of predation or parasitism 4.7.3 Risk of desiccation 4.7.4 Missed opportunities for feeding or reproduction 4.7.5 Interactions between different costs 4.8 Endothermic thermoregulation 4.8.1 The evolutionary origins of endothermy 4.8.2 Optimal thermoregulation by endotherms 4.9 Conclusions

88 88 91 91 96 98 102 105 105 106 107 111 111 112 116 116 118 119 119 122 125

5

Thermal Acclimation 5.1 Patterns of thermal acclimation 5.2 The beneﬁcial acclimation hypothesis 5.2.1 Developmental acclimation 5.2.2 Reversible acclimation 5.2.3 Beyond the beneﬁcial acclimation hypothesis

126 126 127 127 131 135

CONTENTS

5.3

5.4

5.5

5.6

5.7 5.8

Costs of thermal acclimation 5.3.1 Costs of energetic demands 5.3.2 Costs of time lags 5.3.3 Interaction between costs Optimal acclimation of performance curves 5.4.1 Optimal developmental acclimation 5.4.2 Optimal reversible acclimation 5.4.3 Relaxing assumptions about ﬁtness Evidence of optimal acclimation 5.5.1 Does the thermal optimum acclimate more than the performance breadth? 5.5.2 Do organisms from variable environments acclimate more than organisms from stable environments? Constraints on the evolution of acclimation 5.6.1 Genetic variance and covariance 5.6.2 Gene ﬂow Toward ecological relevance Conclusions

xi

135 136 139 139 140 141 143 145 146 146 146 149 149 152 154 155

6

Temperature and the Life History 6.1 The link between performance and the life history 6.2 General patterns of age and size at maturity 6.2.1 Thermal plasticity of age and size at maturity 6.2.2 Thermal clines in age and size at maturity 6.2.3 Experimental evolution of age and size at maturity 6.3 Optimal reaction norms for age and size at maturity 6.3.1 A comparison of two modeling approaches 6.3.2 Thermal effects on juvenile mortality 6.3.3 Thermal constraints on maximal body size 6.3.4 Thermal effects on population growth 6.3.5 A synergy of evolutionary mechanisms 6.4 General patterns of reproductive allocation 6.4.1 Thermal plasticity of offspring size 6.4.2 Thermal clines in offspring size 6.4.3 Experimental evolution of offspring size 6.5 Optimal size and number of offspring 6.5.1 Direct effect of temperature on the optimal offspring size 6.5.2 Indirect effects of temperature on the optimal offspring size 6.5.3 Teasing apart direct and indirect effects on reproductive allocation 6.6 Optimal variation in offspring size 6.7 Conclusions

157 157 158 158 159 161 162 163 164 166 169 170 171 171 171 172 174 174 176 177 178 179

7

Thermal Coadaptation 7.1 Traits interact to determine ﬁtness 7.2 Coadaptation of thermal sensitivity and thermal acclimation 7.3 Coadaptation of thermal physiology and thermoregulatory behavior 7.3.1 Mechanisms favoring a mismatch between preferred temperatures and thermal optima 7.3.2 Predicting coadapted phenotypes 7.4 Coadaptation of thermoregulatory behavior, thermal physiology, and life history 7.5 Constraints on coadaptation 7.6 Conclusions

181 181 182 186 189 193 195 196 196

xii

CONTENTS

8

Thermal Games 8.1 Filling the ecological vacuum 8.2 Approaches to the study of frequency-dependent selection 8.3 Optimal thermoregulation in an evolutionary game 8.3.1 Competition during thermoregulation 8.3.2 Predation during thermoregulation 8.3.3 Relaxing assumptions of simple models 8.4 Optimal performance curves in an evolutionary game 8.4.1 The coevolution of thermal optima between species 8.4.2 The coevolution of thermal breadths between species 8.4.3 Gene ﬂow and the coevolution of thermal optima 8.5 Life-history evolution in a thermal game 8.6 Conclusions

199 199 200 200 201 203 205 207 207 210 211 212 213

9

Adaptation to Anthropogenic Climate Change 9.1 Recent patterns of climate change 9.1.1 Global change 9.1.2 Regional change 9.1.3 Local change 9.2 Observed responses to recent thermal change 9.2.1 Shifts in phenology 9.2.2 Shifts in geographic ranges 9.2.3 Disruption of ecological interactions 9.2.4 Changes in primary productivity 9.3 Predicting ecological responses to global warming 9.3.1 Correlative versus mechanistic models 9.3.2 Mechanistic models of responses to environmental warming 9.3.3 Predicting differential responses of populations and species 9.4 Adaptation to directional thermal change 9.4.1 Adaptation of thermoregulation 9.4.2 Adaptation of the thermal optimum 9.4.3 Adaptation of the performance breadth 9.5 Thermal games in a warming world 9.6 Evolutionary consequences of gene ﬂow in a warming world 9.6.1 Spatially heterogeneous warming can reduce the ﬂow of maladapted genotypes 9.6.2 Spatially heterogeneous warming can increase the ﬂow of preadapted genotypes 9.7 Conclusions

214 214 214 215 216 216 216 217 217 218 218 219 220 223 224 227 229 231 232 233 233 235 236

References

238

Author Index

277

Species Index

283

Subject Index

285

CHAPTER 1

Evolutionary Thermal Biology

1.1 The challenge of evolutionary thermal biology Of the hundreds of variables that can shape the phenotype of an organism, temperature has undoubtedly captured more than its share of attention.1 Biologists have linked temperature to everything from temporal patterns of growth, survival, and reproduction (Angilletta et al. 2002a; Huey and Stevenson 1979; Savage et al. 2004) to spatial patterns of body size, population density, and species diversity (Angilletta et al. 2004a; Brown et al. 2004; Wiens et al. 2006). Unlike many other variables that concern biologists, temperature is not just a property of life; it is a property of matter. Nothing escapes its control. Still, the same change in temperature will not affect all organisms equally. Moreover, temperature does not affect the same organism equally at all stages of the life cycle. These phenomena raise serious questions for biologists. To what extent does this phenotypic variation reﬂect adaptation to temperature? And what

1

Thermal biologists often make sweeping statements of this nature, but what evidence do we have to back up this view? I became curious while working on this book, so I performed a simple analysis using the ISI Web of Science. Twelve other wellstudied variables were chosen as key words, representing both abiotic and biotic factors: competitors (or competition), energy, food, light, nitrogen, oxygen, parasites (or parasitism), pH, predators (or predation), pressure, salinity, and water. To limit the number of extraneous references, I used the term “biology” in conjunction with the environmental variable (e.g., “temperature and biology”); I have no reason to think this approach would bias the outcome. Since 1945, more than 3568 papers have used the terms “temperature” and “biology.” Temperature ranks third on the list, behind water (4369) and light (3723) but well ahead of food (3029), pressure (1920), parasites (1917), oxygen (1769), predators (1621), and other variables (<1500 each).

processes facilitate or constrain this thermal adaptation? To answer these questions, we must ﬁrst identify the speciﬁc phenotypes that we seek to explain. Strategies for coping with thermal heterogeneity deﬁne a continuum of two dimensions (Fig. 1.1). The ﬁrst dimension describes the degree to which an organism’s performance depends on its temperature (thermosensitivity or thermal sensitivity), ranging from organisms whose performance depends strongly on temperature (thermal specialists) to organisms that perform well over a broad range of temperatures (thermal generalists). The second dimension describes the degree to which an organism regulates its temperature (thermoregulation), ranging from organisms that maintain a nearly constant body temperature (perfect thermoregulators) to organisms that conform to their environmental temperature (perfect thermoconformers). All organisms fall somewhere within the continuum bounded by these axes. For example, most mammals and birds are thermal specialists that thermoregulate precisely (Cossins and Bowler 1987). Certain ﬁsh, such as those that live in the intertidal zone, are thermal generalists that can undergo dramatic ﬂuctuations in temperature during a single day (Somero et al. 1996). The position of an organism along this continuum can change over time. In fact, an individual can be a specialist during speciﬁc seasons but a generalist over the course of a year. Thus, strategies for coping with thermal heterogeneity include irreversible and reversible forms of phenotypic plasticity. These strategies determine the life histories of organisms by limiting their acquisition and allocation of resources. The relationships among thermosensitivity, thermoregulation, and the life history govern the course of adaptation in variable environments. 1

T H E R M A L A D A P TAT I O N

Th Sp er ec m ial os ist en s

Ge ne ra l

iti

vi ty

is t

2

1 3

2

7b

7a 4

P ther erfect mor egu lato r

The

rmo

regu

5

latio

n

Figure 1.1 Strategies of coping with thermal heterogeneity include different combinations of thermosensitivity and thermoregulation. The shaded regions depict the strategic sets of hypothetical species; intraspeciﬁc variation determines the areas of these sets. A strategic set can change seasonally or ontogenetically, as exempliﬁed by the arrow connecting two sets for the same species (7a and 7b).

6

ther Perfect moc onfo rme r

Although a vast and venerable literature documents the scope of these strategies, we still do not understand why certain species exhibit certain phenotypes. Why do some organisms function over a wide range of temperature when other organisms do not? Why do some organisms adjust their thermosensitivity in response to extreme or ﬂuctuating temperatures when other organisms do not? I think most biologists would agree that the answers to these questions involve adaptation. Still, we would be hard pressed to provide rigorous answers at this time. How can we know so much about the thermal biology of organisms, yet know so little about its evolution? I believe progress in evolutionary thermal biology has lagged behind progress in other evolutionary disciplines because of the paucity and simplicity of our theoretical models.2 In fact, few investigations have been directly motivated by quantitative models. And even fewer have been motivated by the desire to distinguish between two or more competing models. In the literature on thermal biology, empirical descriptions of patterns probably outnumber quantitative models of evolution by more than

100 to one. This situation partly resulted from historical factors. The study of thermal biology has its roots in ecological physiology (Bennett 1987a), which has always been a relatively empirical discipline.3 Only recently have ecological physiologists seriously shifted their attention from questions concerning what, when, and how to ponder the question of why. This conceptual transition catalyzed the origin of evolutionary physiology, which has ﬂourished in recent decades (Burggren and Bemis 1990; Feder et al. 2000; Garland and Carter 1994; Magnum and Hochachaka 1998). Despite this shift in focus, empirical observations have continued to accumulate far more rapidly than have theoretical ideas. Today, the limited number of quantitative models prevents effective application of the scientiﬁc method to evolutionary thermal biology (see Hilborn and Mangel 1997). In this book, we shall draw on models from the general discipline of evolutionary biology to gain rigorous insights about thermal adaptation. First, we must construct a framework for understanding

3

2

Although you may view this statement as a harsh criticism, you should consider the role of quantitative models in a comparable discipline to appreciate my point. For example, biologists who study how organisms adapt to variation in the quantity and quality of food have made tremendous progress over the past few decades, largely because of the development and application of theoretical models (reviewed by Kamil et al. 1987; Stephens et al. 2007; Stephens and Krebs 1986).

As an aside, I believe the controversy surrounding the metabolic theory of ecology stems from its unique attempt to explain a wealth of physiological data with a simple mathematical model (see Brown et al. 2004). Despite the need for caution when invoking such unifying theories, opponents of the metabolic theory seem oblivious to the fact that no alternative model currently offers greater predictive power. Perhaps the strong reaction to this theory reﬂects the mindset of ecological physiologists as much as it reﬂects limitations of the models.

E V O L U T I O N A RY T H E R M A L B I O L O G Y

thermal adaptation that integrates current theoretical paradigms and modern empirical approaches. Then, we can use this framework to explore the structure of current theory and the level of empirical support; this exploration will begin with models that consider mechanisms operating at the organismal level and build up to models that account for the interactions among organisms. Along the way, we shall draw on biochemical and genetic evidence to validate key assumptions. By synthesizing models of behavioral, physiological, and life-historical evolution, we should advance toward a general theory of thermal adaptation. By the end of the book, we will have addressed many questions concerning thermal adaptation: 1. How do organisms experience thermal heterogeneity in their environment? Do different organisms perceive the same thermal environment differently? If so, how do we deﬁne the operative environment experienced by a particular organism? (Chapter 2) 2. Why do some organisms function over a wide range of temperatures, while others function only within a relatively narrow range? (Chapter 3) 3. Why do some organisms regulate their body temperature within narrow limits, while others conform to environmental temperatures? (Chapter 4) 4. Why do some organisms possess a capacity to reversibly or irreversibly alter their sensitivity to temperature, while other organisms do not? (Chapter 5) 5. Despite the diversity of behavioral and physiological responses to temperature, why do general relationships exist between environmental temperature and the life history? (Chapter 6) 6. Can we answer the preceding questions independently, or must we consider the evolutionary interactions among thermosensitivity, thermoregulation, and the life history? (Chapter 7) 7. Can we understand thermal adaptation without considering the interactions among individuals of one or more species? (Chapter 8) 8. Can a theory of thermal adaptation help us to predict the biological consequences of anthropogenic climate change? (Chapter 9) Any effort to address this many questions in a single book can succeed only with some focus. This

3

focus could take one of several forms—taxonomic, methodological, or conceptual. I have chosen a conceptual focus: the thermal reaction norm. All of the models presented in this book describe the evolution of some type of thermal reaction norm. Therefore, we must start by deﬁning this fundamental concept and recognizing its strengths and limitations.

1.2 Thermal reaction norms Thermal responses at all levels of biological organization are best described as phenotypic plasticity. The study of phenotypic plasticity has a long and complex history (Sarkar 2004; Schlichting and Pigliucci 1998), but most research in this area has been guided by the concept of a reaction norm. A reaction norm describes the relationship between a continuous environmental variable and a continuous phenotypic variable (see reviews by Gotthard and Nylin 1995; Schlichting and Pigliucci 1998; Stearns 1989; West-Eberhard 2003). A thermal reaction norm represents a special case in which the environmental variable is temperature. In practice, any aspect of behavior, physiology, morphology, or life history can be viewed in the context of a thermal reaction norm (DeWitt and Scheiner 2004; Stearns 1989). Although biologists deﬁne plasticity in numerous ways (Debat and David 2001), I deﬁne plasticity as the derivative of a reaction norm (Fig. 1.2). For a linear reaction norm, plasticity simply equals the slope of the function. For a nonlinear reaction norm, the concept of plasticity becomes muddled as the slope of the function depends on the temperature of interest. A trait unaffected by temperature has no plasticity, but this trait still has a reaction norm; the slope of this reaction norm equals zero and the intercept equals the mean value of the trait. Given this deﬁnition of plasticity, we can compare plasticities among traits or taxa. Moreover, we can quantify the evolution of plasticity for speciﬁc traits. The trait of interest affects the way we conceptualize and quantify a thermal reaction norm. Some traits, such as growth rate, change repeatedly in response to temperature (labile traits). Other traits, such as the number of body segments, become ﬁxed during development (ﬁxed traits). For a labile trait,

4

T H E R M A L A D A P TAT I O N

(a)

(a) Cold specialist

Low plasticity

Phenotype

Phenotype

No plasticity

Heat specialist

Generalist

High plasticity Body temperature High Low plasticity plasticity

(b) Parameter describing thermal sensitivity

High plasticity

Phenotype

(b)

Body temperature Figure 1.2 The derivative of a reaction norm quantiﬁes the thermal plasticity. For linear reaction norms (a), plasticity remains constant over a broad range of temperatures, but can vary among genotypes. For nonlinear reaction norms (b), the plasticity expressed by a single genotype depends on the range of temperatures.

we can measure the change in the trait of an individual exposed to a range of temperatures (Fig. 1.3a). Such studies are usually designed to avoid confounding the temperature and time of measurement (e.g., randomizing the order of measurements solves this problem). Reaction norms for labile traits often depend on the developmental environment of the organism, such as its developmental temperature. In other words, thermal plasticity itself can undergo plasticity! To quantify this “plasticity of plasticity,” we can examine the relationship between the developmental temperature and the thermal sensitivity after development (Fig. 1.3b). For a ﬁxed trait, a reaction norm must be estimated from measures of relatives raised at different temperatures. Using a group of distantly related individuals would yield a generalized reaction norm (Sarkar and Fuller 2003), which includes phenotypic variation among genotypes. Using clones or siblings would yield a more accurate reaction norm. Reaction norms

Developmental temperature Figure 1.3 Thermal reaction norms characterize many different kinds of traits. (a) For labile traits, instantaneous (or short-term) ﬂuctuations in body temperature affect the phenotype. (b) Thermal reaction norms can change via acclimation; such changes are described as plasticity in the parameters that deﬁne the reaction norm shown in (a).

for ﬁxed traits usually describe the inﬂuence of the mean developmental temperature. If thermal ﬂuctuations are important, we can construct a multivariate reaction norm that describes the phenotype as a function of the mean and the variance of temperature (Fig. 1.4). This diversity of applications underscores the conceptual value of the reaction norm. How can we know whether thermal reaction norms were produced by adaptation? To acquire such knowledge, we must construct predictive models of evolution and apply these models to the study of thermal plasticity. In other words, we must develop a theory of thermal adaptation. With this goal in mind, the remainder of this chapter focuses on the role of theory in evolutionary thermal biology.

Phenotype

E V O L U T I O N A RY T H E R M A L B I O L O G Y

40 Va ria 30 nce of te

40 30 10 re

10

an

Me

ure

rat

20

20 mp era tu

e mp

te

Figure 1.4 A multivariate reaction norm can describe the inﬂuence of the mean and the variance of developmental temperature on a ﬁxed trait. In this hypothetical example, a univariate reaction norm would give an incomplete view of thermal plasticity because of the interaction between the mean and the variance of temperature.

1.3 The role of theory In his book Evolution in Changing Environments, Richard Levins (1968) opened with a visionary discussion of the role of theory in biology. He predicted that theoretical biology, an emerging discipline at the time, would soon disappear. I interpret his argument as follows. Science has always been a process of building and evaluating models of how nature works. But some biologists had lost sight of this reality, which eventually fueled a movement to bring theoretical direction to empirical research. Levins reasoned that, as more biologists accepted the need for theory, the distinction between theoretical biology and empirical biology would dissolve. In Levins’ own words, “once it [theoretical biology] has made its points part of the thinking of biology in general, it will loose [sic] its identity as a distinct ﬁeld” (p. 4). Sadly, Levins’ admirable vision did not come to pass. Even today, a fairly wide rift separates biologists who conduct theoretical research from those who conduct empirical research. I shall voice an extreme view (hopefully, one that Levins would espouse): scientiﬁc understanding advances most rapidly when researchers combine theory and empiricism. At the very least, we need a reciprocal exchange of ideas between theorists and empiricists;

5

theorists should model processes in light of previous observations and empiricists should collect data that address current models. My motivation for writing this book was to bridge the gap between theorists and empiricists who study thermal adaptation. To achieve this goal, I must convince you that progress in evolutionary thermal biology hinges on this connection. As a ﬁrst step, I should elaborate my view on the interplay between models and data. A theory consists of a group of models describing related phenomena (e.g., foraging theory). Each model describes a set of mechanisms proposed to explain the phenomena. These models inspire the hypotheses and experiments that capture the attention, time, and energy of biologists. Yet, different forms of modeling exist and not all of these forms can offer the same degree of insight. Indeed, models differ greatly in the clarity of their assumptions and the precision of their predictions. Most models can be characterized as verbal, graphical, or mathematical. A verbal model is a textual description of the modeler’s ideas. Generally, the assumptions of a verbal model are implicit. Furthermore, its predictions would be qualitative and could be inaccurate if the modeler has not processed his assumptions carefully. Because of this imprecision, we cannot easily evaluate and reﬁne verbal models. A graphical model depicts a set of relationships among variables plotted in Euclidean space. Although the assumptions of a graphical model are also implicit, one can infer them by analyzing the graph. The predictions of the model can be qualitative or quantitative depending on whether the modeler has speciﬁed numerical values of the variables. Finally, a mathematical model embodies the assumptions of the modeler in an algebraic form. The relationships among variables and the units of measurement are deﬁned explicitly. As such, mathematical models embody clear assumptions and deliver precise predictions. In this book, I focus primarily on mathematical models rather than verbal or graphical models. This focus does not imply that mathematical models provide the best representations of truth; in fact, they do not (see Lakoff and Núñez 2000). All models, including mathematical ones, represent caricatures of nature more than illustrations of truth (Hilborn and Mangel 1997). But the explicit assumptions

6

T H E R M A L A D A P TAT I O N

of mathematical models enable other biologists to (1) verify the validity of those assumptions and (2) generate competing models by altering the assumptions. Furthermore, mathematical models generate more comprehensive and more precise predictions, which facilitate distinctions among competing models that make qualitatively identical predictions under certain conditions. We can evaluate a model in one of two ways. First, we can validate the critical assumptions of the model. Acritical assumption determines the model’s qualitative behavior. Relaxing some assumptions will cause only quantitative changes in behavior, but a critical assumption underlies some unique prediction. If this assumption proves false, the prediction would also be false. Second, we can compare independent observations with the predictions of the model. By independent observations, I mean data that were not used to formulate or parameterize the model. Throughout the book, I adopt both approaches to evaluate models. Rather than evaluate one model at a time, we gain deeper insights by directly comparing the performances of two or more models (Chamberlin 1890). In the traditional statistical paradigm, the prediction of a single model is pitted against a null hypothesis, which generally states that no pattern or effect should be observed. Presumably, rejection of the null hypothesis implies support for the model (also known as afﬁrming the consequent). Unfortunately, many other models could have made a prediction similar to that of the model in question. Thus, competition between an evolutionary model and a null model provides only weak support for a theoretical idea (Anderson et al. 2000; Oreskes et al. 1994). Competition between two or more evolutionary models always yields a more fruitful outcome. For example, competition between mutually exclusive models eliminates some ideas from the set of possibilities. In this way, each experiment brings us closer to identifying the best available model (Platt 1964). When a set of models describes mutually inclusive mechanisms, we can still infer the relative importance of each mechanism (Quinn and Dunham 1983). Today, we can choose from several statistical paradigms, such as information theory (Burnham and Anderson 2002) or Bayesian theory (Ellison 2004), to directly compare the performance

of two or more models (reviewed by Johnson and Omland 2004). These paradigms enable us to formally rank a set of models by their ability to describe the data. But even an informal comparison of models is better than no comparison at all. Therefore, let us examine the kinds of theoretical models available for the study of thermal adaptation.

1.4 Theoretical approaches to evolutionary thermal biology Three distinct modeling approaches will appear throughout this book: optimality models, quantitative genetic models, and allelic models (Berrigan and Scheiner 2004; Scheiner 1993). Each of these approaches possesses unique strengths, yet each also imposes some limitations. In this section, I outline the basic features of these three approaches. While doing so, I highlight the way that each approach informs us about thermal adaptation.

1.4.1 Optimality models An optimality model describes the phenotype that maximizes ﬁtness in a given environment (Box 1.1). This goal is accomplished by combining two mathematical functions: (1) a phenotypic function that deﬁnes the set of possible phenotypes and (2) a ﬁtness function that relates these phenotypes to ﬁtness in given environments (Maynard Smith 1978; Mitchell and Valone 1990). Combining the phenotypic and ﬁtness functions enables one to map the ﬁtness landscape—a graphical depiction of the relationship between potential phenotypes and their ﬁtness. A ﬁtness landscape tells us the direction and strength of natural selection given the current phenotype. Under certain conditions (described below), natural selection would ultimately yield the phenotype that maximizes ﬁtness (i.e., the optimal phenotype). The heart of an optimality model is a set of assumptions about the tradeoffs that constrain the phenotype (Mitchell and Valone 1990), sometimes referred to as functional constraints (Arnold 1992). Biologists often consider one of two classes of tradeoffs: allocation tradeoffs or acquisition tradeoffs. An allocation tradeoff occurs when an increment in resources allocated to one function results in a decrement in resources allocated to other functions (Heino

E V O L U T I O N A RY T H E R M A L B I O L O G Y

7

Box 1.1 Types of optimality models Optimality models are divided into three classes: static, dynamic, and frequency dependent (Maynard Smith 1978). Below, these classes are described in order of increasing complexity. Static optimization Static models optimize one or more variables in an environment that does not change over time. Do not equate static with deterministic; the optimal strategy could depend on probabilities (e.g., the daily probability of surviving), but these probabilities would remain constant. Static optimization requires the use of differential calculus to determine the phenotypic value that maximizes ﬁtness (or some proxy for ﬁtness). Dynamic optimization Dynamic models optimize one or more variables in an environment (internal or external) that changes over time.

and Kaitala 1999; Kozłowski 1992). An acquisition tradeoff occurs when an increase in the duration or intensity of foraging results in an increase in the probability of being injured or killed (Lima and Dill 1990). Support for a model indirectly supports the importance of the speciﬁed tradeoffs. When modeling thermal reaction norms, we must consider additional tradeoffs. Models of optimal reaction norms deﬁne the degree of plasticity that maximizes ﬁtness in a variable environment. Because all optimality models link environments, phenotypes, and ﬁtness, any model can predict the optimal reaction norm when two conditions hold: (1) the phenotype expressed in one environment does not constrain the phenotypes expressed in other environments and (2) plasticity imposes no decrease in survivorship or loss of energy. We would merely predict the optimal phenotype in each of several environments and connect the points to deﬁne a reaction norm. But this approach ignores potentially important constraints on the evolution of thermal reaction norms. For example, a specialist–generalist tradeoff occurs when a phenotype that improves ﬁtness at one temperature also reduces ﬁtness at

In a dynamic model, the organism makes sequential decisions. These decisions cause changes in variables that describe the state of the environment or the organism. For example, a dynamic optimization of reproduction involves rates of energy intake and expenditure that increase with body size (a phenotypic state variable). Dynamic optimization usually requires the application of optimal control theory, such as dynamic programming. Frequency-dependent optimization When the ﬁtness of a strategy depends on the strategies adopted by other individuals (i.e., frequency-dependent selection), one must take a game theoretic approach to optimization (Maynard Smith 1982; Nowak and Sigmund 2004). In game theory, the optimal strategy cannot be invaded by any alternative strategy—hence it is called an evolutionarily stable strategy. As we shall see in Chapter 8, the evolutionarily stable strategy often deviates from the optimal strategy under frequency-independent selection.

another temperature (Levins 1968; Palaima 2007). Additionally, various costs of plasticity can affect the optimal reaction norm (reviewed by DeWitt et al. 1998; Schlichting and Pigliucci 1998). Hence, the optimal reaction norm in a variable environment often deviates from the optimal phenotype in a constant environment. Optimality models comprise two kinds of assumptions: hard-core and strategic assumptions (Mitchell and Valone 1990). The hard-core assumptions are the foundation of the optimization research program and thus are the same for all models: 1. Mutation generates a small number of new phenotypes each generation. 2. Different phenotypes yield different ﬁtnesses. 3. Populations of phenotypes typically reside near a stable equilibrium. 4. The relative ﬁtnesses of phenotypes near the stable equilibrium are not inﬂuenced by the way these phenotypes are transmitted genetically. The strategic assumptions depend on the speciﬁc problem under investigation and thus differ among models. Strategic assumptions are related to either

8

T H E R M A L A D A P TAT I O N

(1) the constraints on the phenotype or (2) the way the phenotype determines ﬁtness. As an example of the former, many optimality models of thermal adaptation assume that organisms cannot enhance function at high temperatures without sacriﬁcing function at low temperatures (see Chapter 3). Other optimality models assume that organisms possess a ﬁxed quantity of energy at any time (see Chapter 6). Either of these assumptions would constrain the set of possible phenotypes. As an example of the latter, an optimality model could describe ﬁtness as the product of survivorship and fecundity. Alternatively, one might assume survivorship remains constant and fecundity increases with body size; hence, size at maturity would be considered an adequate proxy for ﬁtness. Strategic assumptions usually determine the qualitative behavior of the model. The failure of an optimality model to predict the observed phenotypes indicates a violation of at least one strategic or hard-core assumption. First, the hypothesized tradeoffs might be incorrect or incomplete. Second, the mapping between the phenotype and ﬁtness might be incorrect, or ﬁtness might be deﬁned incorrectly. Finally, we might question the hard-core assumptions, which are necessary for selection to maintain the optimal phenotype. Genetic drift impedes adaptation in small populations (Holt 1990), and mutation drives phenotypes away from the optimum when natural selection is weak (Gilchrist and Kingsolver 2001). Persistent gene ﬂow along an environmental cline swamps a population with deleterious genes, despite strong selection in local environments (Lenormand 2002). Moreover, natural selection produces few evolutionary changes when phenotypic variation lacks a genetic basis (Arnold 1992). In the end, we might have to model these additional factors that constrain thermal adaptation. Let us now consider some approaches that emphasize the role of genetic constraints during adaptation.

1.4.2 Quantitative genetic models Quantitative genetic models enable us to relax some of the hard-core assumptions of optimality models and examine the evolutionary dynamics of traits. In quantitative genetic models, traits are determined

by many genes, each of which has a small quantitative effect on the phenotype (Gillespie 2004; Rice 2004). This theory partitions phenotypic variance into genetic and environmental components. In his seminal book, Ronald Fisher (1930) showed that the mean ﬁtness of a population increases in proportion to the genetic component of phenotypic variation. In particular, he distinguished between the additive and nonadditive effects of alleles. Only the additive portion of genetic variation contributes to the evolution of ﬁtness; the nonadditive portion stems from interactions among genes (dominance and epistasis), which are altered by recombination. Thus, the phenotypic variance (VP ) must be partitioned as VP = VG + VE ,

(1.1)

where VG equals the variance attributable to genetic effects and VE equals the variance attributable to environmental effects. If we assume only additive genetic effects exist, the heritability of the phenotype (h2 ) equals the proportion of variance caused by genetic effects: h2 =

VG VG = . VP VG + V E

(1.2)

Thus, the value of a phenotype depends not only on genes but also on environmental factors, such as temperature. When the effect of the environment on the phenotype depends on the genotype, a genotype-by-environment interaction exists. In such cases, a better model for partitioning the phenotypic variance would be VP = VG + VE + (VG×E ) + VError ,

(1.3)

where VG×E equals the variance caused by the interaction between genes and the environment—i.e., the genetic variance in phenotypic plasticity. Statistical methods can partition the components of variance and estimate the magnitude of the genotype-byenvironment interaction (Lynch and Walsh 1998). With this framework alone, researchers have quantiﬁed genetic variation in thermal reaction norms (e.g., Scheiner and Lyman 1991), but predicting their evolution requires (1) a more precise means of estimating genetic variances and (2) a signiﬁcant advance in modeling selection. Both advances were made during the development of multivariate selection theory (Via and Lande 1985).

E V O L U T I O N A RY T H E R M A L B I O L O G Y

Multivariate selection theory enables us to predict the evolution of thermal reaction norms under genetic constraints (reviewed by Arnold 1994). At the root of this theory lies a model of the simultaneous evolution of two or more traits (Lande 1979; Lande and Arnold 1983). This model describes change in the phenotype as a function of selective pressure (estimated by selection gradients) and genetic variation (estimated by genetic variances and covariances): Z = βG,

(1.4)

where Z equals a vector of changes in phenotypic values, β equals a vector of selection gradients, and G equals a matrix of genetic variances and covariances. For example, the evolution of three traits would be represented by      Z1 β1 G11 G12 G13            Z2  =  β2   G21 G22 G23  , (1.5)      Z3

β3

G31

G32

G33

where Zi equals the change in trait i, β i equals the selection gradient for trait i, Gii equals the genetic variance of trait i, and Gij equals the genetic covariance between traits i and j. A selection gradient equals the partial derivative of relative ﬁtness with respect to the phenotype; in other words, a selection gradient measures the change in ﬁtness that results from a very small change in the phenotype. In the special case of directional selection, the selection gradient equals a constant. In the case of stabilizing or disruptive selection, however, the selection gradient varies with the phenotypic value. The genetic variance in eqn 1.5 equals the additive genetic variance described by Fisher (VG of eqn 1.2); thus, the rate of adaptation will be proportional to the genetic variance. Many factors can limit the genetic variance of a trait (Table 1.1). The genetic covariance between two traits determines how the selection of one trait will inﬂuence the evolution of another trait. For example, when the genetic covariance is negative (Gij < 0), selecting a larger phenotypic value of trait j will decrease the phenotypic value of trait i. Genetic covariances result from several mechanisms, including linkage disequilibrium, pleiotropy, and developmental modularity (Lynch and Walsh 1998; West-Eberhard 2003).

9

Table 1.1 Factors that reduce the genetic variance of a speciﬁc trait 1. Persistently small population leads to loss of variation by genetic drift 2. Mutational effects are skewed by mechanistic/physiological constraints 3. Small target of mutation lowers the rate of mutation 4. Mutation rates are unusually low 5. New allelic associations are not frequently formed by recombination 6. The phenotype is canalized during development despite mutations 7. The trait is genetically correlated with another trait under selection 8. The same genotype has the highest ﬁtness in all environments 9. Continuous directional selection leads to ﬁxation of favored alleles Modiﬁed with permission from Blows and Hoffman (2005).

To use multivariate selection theory, we must represent a thermal reaction norm as a set of traits with genetic variances and covariances. Several approaches exist for translating a reaction norm into a set of traits: (1) the character-based approach, (2) the parameter-based approach, (3) the characteristic value-based approach, and (4) the function-based approach. Although these approaches should produce similar results (de Jong 1995), they involve different mechanics. In the character-based approach, one treats the same trait measured at multiple temperatures as equivalent to multiple traits measured at one temperature (Falconer 1952; Via 1987; Via and Lande 1985). Each element of the G matrix becomes the additive genetic variance of the phenotype expressed at a temperature (Gii ) or the additive genetic covariance between the phenotypes expressed at different temperatures (Gij ). The vector of selection gradients determines the strengths of selection acting on a single trait at different temperatures. In the parameter-based approach, one describes the reaction norm with a mathematical function, and treats parameters of the function as equivalent to traits (Scheiner and Lyman 1991). Although any function can be used, mainly linear and polynomial functions have been explored (David et al. 1997; Gibert et al. 1998). For a linear function, the slope represents the plasticity of the phenotype and the intercept represents the mean of the phenotype. As in the trait-based approach, a genetic variance–covariance matrix is used in conjunction

10

T H E R M A L A D A P TAT I O N

with selection gradients to model the evolution of the reaction norm. Selection gradients for the parameter-based approach apply to parameters of the function. In the characteristic value-based approach, one describes the reaction norm with a mathematical function and then derives characteristic values from this function to serve as traits (Gibert et al. 1998). A subtle but important distinction exists between a parameter and a characteristic value. Parameters, such as a slope and intercept, are estimated by ﬁtting a model to data. Although some parameters have obvious biological meanings, other parameters do not. We can replace biologically meaningless parameters with characteristic values of the function, such as the phenotype at a particular temperature or the mean phenotype over a range of temperatures. The evolution of these characteristic values could be predicted in the same way that one predicts the evolution of the parameters. In the function-based approach, one treats the reaction norm as a trait with an inﬁnite number of dimensions (Gomulkiewicz and Kirkpatrick 1992). In other words, the thermal reaction norm describes the phenotype at an inﬁnite number of temperatures. The inﬁnite-dimensional approach relies on nonparametric statistics, which have the advantage of relaxing some restrictive assumptions of the other approaches. An inﬁnite-dimensional analysis yields principal components that describe the genetic variance and covariance of the reaction norm. These estimates of genetic variance and covariance can be used to model the response to selection (Kingsolver et al. 2001a). Regardless of which approach we choose, applying a quantitative genetic model poses unique challenges and requires new assumptions. First, we must know the selection gradients corresponding to the traits that represent the reaction norm. We can obtain these gradients from the ﬁtness landscape of an optimality model (see Section 1.4.1) or estimate them empirically (see Section 1.5.1). Second, we must know the genetic variances and covariances associated with these traits. We can estimate these parameters through breeding experiments or artiﬁcial selection (Box 1.2). Regardless of how we obtain genetic variances and covariances, most

quantitative genetic models assume these parameters remain constant during evolution. In reality epistasis, selection, recombination, and migration change the genetic variances and covariances of real traits (Archer et al. 2003; Hansen 2006; Pfrender and Lynch 2000; Roff and Mousseau 1999; reviewed by Steppan et al. 2002). Other assumptions arise from the statistical methods used to estimate genetic variances and covariances (Table 1.3). Despite these assumptions, quantitative genetic models provide a valuable complement to optimality models (Roff 1994), as we shall see in later chapters.

1.4.3 Allelic models As the name implies, an allelic model describes the genes that shape the phenotype. Thus, allelic models pry open the black box of genes created by quantitative genetic models. The classic analytical models of one or two loci (reviewed by Gillespie 2004) have given way to more complex models explored through computer simulations of genetic algorithms (reviewed by Mitchell and Taylor 1999; Schoenauer and Michalewicz 1997). The details vary greatly from model to model, but the allelic models described in this book all share some general properties. These models focus on a population of diploid organisms, with multiple alleles per locus. Each allele has an additive effect on the genetic value of a trait. For our purposes, the trait will represent some parameter that describes a thermal reaction norm, as in the parameter-based approach of quantitative genetics (see Section 1.4.2). The phenotypic value of this parameter results from the sum of allelic effects and an environmental effect. Because the mean environmental effect equals zero, the mean phenotypic value of the trait equals the mean genetic value. The phenotypic variance equals the sum of the additive genetic variance and the environmental variance (see eqn 1.1). Initially, a population of genotypes is formed by randomly assigning a set of alleles to each individual. Then, within each generation, individuals survive and reproduce according to their phenotype. Upon reproduction, some probability exists that the genetic value of an allele will change by mutation; most mutations cause very small changes

E V O L U T I O N A RY T H E R M A L B I O L O G Y

Box 1.2 Estimating parameters of quantitative genetic models Quantitative genetic models of thermal adaptation require estimates of genetic variances and covariances. Estimating these parameters for a reaction norm requires a fairly complex experimental design (Scheiner 2002). For labile traits, one can quantify a thermal reaction norm through repeated measures of individuals. For ﬁxed traits, however, one must estimate the reaction norm using clones or siblings, because the phenotype of an individual cannot be measured at multiple temperatures. Quantitative genetic parameters estimated from siblings should be viewed with caution because microenvironmental variation can inﬂate estimates of genetic variance (Windig et al. 2004). Genetic variances and covariances can be quantiﬁed through breeding experiments or artiﬁcial selection (Falconer 1989; Lynch and Walsh 1998). Breeding experiments Relationships among relatives can be used to estimate genetic variances and covariances. Three kinds of relatives are commonly used: (1) parents and offspring, (2) full siblings, and (3) half-siblings. In each case, males (sires) and females (dams) are randomly mated to produce a set of offspring. These offspring are reared in controlled environments to reduce environmental effects on the phenotype. In the context of thermal adaptation, the trait could be a phenotype measured at a single temperature or a parameter that describes a thermal reaction norm. The additive genetic variance of each trait can be determined by linear regression or analysis of variance (ANOVA). When a trait is measured in both parents and their offspring, a linear regression of the offspring phenotype on the mean parental phenotype reveals the additive genetic variance. Speciﬁcally, the proportion of variance in the offspring phenotype explained by the parental phenotype (r 2 ) approximates the heritability of the

trait (VG /VP ; see eqn 1.2). Alternatively, the phenotypic variance among full or half-sibling families can be used to estimate the additive genetic variance. For a full sibling design, each male is mated randomly to a single female. A nested ANOVA estimates the proportion of phenotypic variance explained by familial identity (sire or dam). Assuming the maternal environment and genetic dominance do not inﬂuence the phenotype, the additive genetic variance equals twice the variance observed among families. When maternal and dominance effects are suspected, a half-sibling design provides a more accurate estimate of additive genetic variance. Each male is mated randomly to several females. A more accurate estimate of the additive genetic variance results because each mother experiences a unique environmental condition and provides a unique genetic complement. A nested ANOVA estimates the proportion of phenotypic variance explained by paternal identity; the additive genetic variance equals four times the variance observed among half-sibling families. By raising individuals from the same family at different temperatures, one can use an ANOVA to estimate the genetic variance of plasticity (Table 1.2). Artiﬁcial selection The response to artiﬁcial selection can also be used to estimate genetic variances and covariances. One starts with a large population of organisms that expresses phenotypic variation in the trait of interest. This population is randomly subdivided to create replicate lines; some lines will experience selection and other lines serve as controls. For each selection line, the researcher selects individuals that have either the highest or the lowest values of the trait. For example, imagine we wished to know the genetic variance of heat tolerance. In some lines, we would select individuals with a heat tolerance greater than the median (i.e., the top 50%). For other lines, we would select

Table 1.2 Interpreting the results of an analysis of variance applied to a breeding experiment conducted at multiple temperatures Factor

Description

Genotype or familial identity (VG ) Temperature (VE ) Genotype × temperature (VG×E ) Error (Ve ) Total (VP )

Genetic variation in the mean phenotype expressed across temperatures Mean thermal plasticity of the phenotype Genetic variation in the thermal plasticity of the phenotype Variation among individuals within genotype at a temperature Phenotypic variance

Modiﬁed with permission from Via (1994). continues

11

12

T H E R M A L A D A P TAT I O N

Box 1.2 Continued

¯z1 = β1 · G11 .

(1.6)

The selection gradient (β 1 ) would depend on the intensity of selection that we imposed (s) and the phenotypic variance (VP ): s , β1 = VP

(1.7)

where s equals the difference between the mean phenotype of the selected population and that of the original population. Substituting eqn 1.7 into eqn 1.6 and solving for the genetic variance yields: ¯z1 VP G11 = . s

(1.8)

Given eqn 1.8, we can easily estimate the genetic variance from the response to our selective regime and the

phenotypic variance. Figure 1.5 illustrates the estimation of ¯z1 /s for fruit ﬂies that were artiﬁcially selected for heat tolerance (Gilchrist and Huey 1999). Furthermore, artiﬁcial selection tells us about the genetic covariance between two traits. For example, imagine we wanted to know whether heat tolerance and cold tolerance covaried genetically. After selecting for greater or lesser heat tolerance, we could use the change in cold tolerance to estimate the genetic covariances (Gij and Gji ): G21 = G12 =

¯z2 VP , s

42

40

38

36 0

5

10

15

25

30

Figure 1.5 Response to selection for heat tolerance (knockdown temperature) in Drosophila melanogaster. Replicate lines were selected for greater or lesser heat tolerance. The slope of each relationship equals ¯zi /s , which is proportional to the genetic variance. Adapted with permission from Macmillan Publishers Ltd: Heredity (Gilchrist and Huey 1999), © 1999.

Assumption

Description

Linearity Additivity

Genetic and environmental effects can be described by a linear statistical model No interactions between genes (epistasis) or between genes and environments (environmental effects on the reaction norm) Genetic and environmental effects are normally distributed Genetic variances and covariances do not change

Modiﬁed with permission from Pigliucci and Schlichting (1997).

20

Cumulative selection differential (s)

Table 1.3 Assumptions of quantitative genetic theory

Normality Constant G matrix

(1.9)

where VP and s apply to the selected trait (heat tolerance) and ¯z2 applies to the correlated trait (cold tolerance). Knockdown temperature (°C)

individuals with a heat tolerance lower than the median (i.e., the bottom 50%). For control lines, we would select individuals randomly with respect to heat tolerance. In all lines, we would randomly mate selected individuals to produce the next generation. The remaining individuals would be eliminated from each line. With each successive generation, the mean heat tolerance of the selected lines should diverge from the mean heat tolerance of the control lines. Using eqn 1.4, we could estimate the genetic variance of heat tolerance (Gii or VG ) from the response to artiﬁcial selection. This equation tells us the response to selection depends on the selection gradients and the genetic variances and covariances of traits. If we assume no trait other than heat tolerance-experienced selection, the response to selection becomes simple:

E V O L U T I O N A RY T H E R M A L B I O L O G Y

in the genetic value but some can have large positive or negative effects (mutational effects are usually drawn from a Gaussian distribution with a mean of zero). A ﬁnite population and independent assortment can lead to genetic drift within these simulations. Over time, the mean phenotype of the population evolves as a consequence of this selection, mutation, and drift. Allelic models confer a major advantage over quantitative genetic models: the genetic variances and covariances can change. Because genes are modeled explicitly, mutation, recombination, drift, and selection change the frequencies of alleles over generations. Ultimately, the trait can evolve to an equilibrial value, such as the evolutionary optimum. Thus, an allelic model provides information about the dynamics and outcome of adaptation. On the downside, allelic models require one to specify not only the ﬁtness landscape but also genomic structure, mutation rate, population size, and mating behavior (Hertz and Kobler 2000). As we shall see, the dynamics of thermal adaptation depend on these genetic and demographic parameters.

13

alleles along with rates of mutation, recombination, and migration, even though these parameters are usually unknown. Thus, no single approach lacks drawbacks. In this book, we shall use all three theoretical approaches to study thermal adaptation. The simplest way to draw on their collective strength is to apply these approaches sequentially. Generally, we shall start by applying an optimality approach to a particular phenomenon. When one optimality model fails, another optimality model may succeed because of its more realistic assumptions. When an optimality approach fails altogether, we shall explore the potential explanations offered by quantitative genetic and allelic models. For example, a lack of additive genetic variance might explain why the optimal phenotype has not evolved. Alternatively, migration between populations in different thermal environments might prevent populations from attaining local optima. This progression from simple to complex approaches will provide greater insights than would a focus on any single approach.

1.4.4 The complementarity of theory All three of the modeling approaches discussed here possess strengths and weaknesses. Optimality models describe ﬁtness landscapes given functional constraints on the phenotype (the phenotypic function) and a hypothetical relationship between the phenotype and ﬁtness (the ﬁtness function). Yet, optimality models assume genetic constraints do not prevent the optimal phenotype from evolving. Furthermore, optimization does not describe the evolutionary dynamics that precede an equilibrial phenotype. In contrast, quantitative genetic and allelic models explicitly account for genetic constraints on phenotypes and describe the dynamics that lead to an evolutionary equilibrium. In these models, however, a ﬁtness landscape must either be assumed or be quantiﬁed empirically (Via et al. 1995). Furthermore, evolutionary dynamics depend largely on the assumptions about genetics. In quantitative genetic models, genetic variances and covariances are assumed to be constant even though they are known to vary (Steppan et al. 2002). In allelic models, one must specify the number of loci and

1.5 Empirical tools of the evolutionary thermal biologist Models help us to understand the diversity of phenotypes observed in nature, but simply formulating a model that describes a previous observation will not advance knowledge. The model must predict unobserved phenomena and new observations must be made to conﬁrm this ability. Testing the kinds of models outlined in Section 1.4 requires us to determine the match between predictions and observations. Throughout this book, we shall rely on three distinct approaches to evaluate models: (1) quantifying selection in natural environments, (2) quantifying evolution during selection experiments, and (3) comparing organisms from different environments. These approaches focus on different temporal scales of evolution, ranging from one generation to hundreds of generations. They also provide different strengths of evidence for adaptation and different levels of generality. Let us brieﬂy review each approach to better understand its strengths and limitations.

14

T H E R M A L A D A P TAT I O N

1.5.1 Quantifying selection Measures of selection in natural environments can be used to evaluate optimality models and parameterize genetic models. Recall that an optimality model deﬁnes the ﬁtness landscape—the relationship between the phenotype and ﬁtness (see Section 1.4.1). This landscape can be described mathematically by parameters called selection gradients. Quantitative genetic and allelic models require knowledge of selection gradients to determine the dynamics of adaptation (see Sections 1.4.2 and 1.4.3). Thus, measures of selection gradients in natural environments can validate critical assumptions of these models. Recently, statistical methods for quantifying selection have undergone rapid development (Box 1.3), enabling one to estimate selection gradients in natural environments. Sometimes, limited variation in thermal reaction norms prevents one from deﬁning the adaptive landscape in sufﬁcient detail. In such cases, artiﬁcial selection can increase the frequency of rare genotypes (Bennett 2003; Conner 2003), thereby expanding the scope of an analysis. Artiﬁcially selecting thermal reaction norms requires a different approach than the one used to select a trait expressed at a single temperature. For labile traits, the reaction norm can be measured by exposing each individual to a range of temperatures. For ﬁxed traits, one must estimate the reaction norm by raising clones or siblings over a range of temperatures. In the latter case, the investigator selects individuals from clonal or familial lines that exhibit the desired reaction norm (Scheiner 2002). In addition to amplifying the frequency of rare genotypes, one can further increase the diversity of phenotypes by engineering ones that do not occur naturally. Both genetic and phenotypic engineering can accomplish this goal. Genetic engineering involves the insertion or deletion of genes to produce changes in the phenotype (see, for example, Tatar 1999). A variety of methods have been developed for genetic engineering in model organisms (reviewed by Tatar 2000). Phenotypic engineering involves either physical or physiological manipulation of the phenotype. In the simplest case, a novel phenotype might be engineered through physical manipulation (Sinervo 1990). More

complex approaches involve exogenous delivery of hormones or neurotransmitters (see, for example, Ketterson and Nolan 1992). Both genetic and phenotypic engineering extend the range of phenotypes for analyses of selection (reviewed by Travis and Reznick 1998).

1.5.2 Experimental evolution Despite the value of short-term measures of selection, observations over longer periods can reveal the dynamics of evolution and possibly even the equilibrial phenotypes. The predictions of evolutionary models have been tested experimentally in artiﬁcial mesocosms and natural environments. In the former case, one maintains replicated populations in controlled thermal environments. In the latter case, one establishes replicated populations in a novel natural environment; for mobile organisms, each population must be physically enclosed to prevent migration. In both cases, one compares the evolutionary responses in experimental populations with those in control populations (Box 1.4). Although both kinds of experiments have the same goal, each has its own advantages and disadvantages. Laboratory experiments directly link environmental temperatures to the evolution of phenotypes (Bennett and Lenski 1999). With a careful experimental design, any difference between the phenotypes of experimental and control lines can be attributed to thermal adaptation (see Box 1.4). However, practical considerations demand organisms with short generations, high fecundity, and tolerance to artiﬁcial conditions (e.g., Drosophila melanogaster or Escherichia coli). Field experiments can reveal novel patterns of thermal adaptation because natural selection depends on the interaction of temperature and other environmental factors. The use of natural environments eliminates the need to care for organisms, but creates difﬁculties when monitoring populations and interpreting results (see Box 1.4). Although these two kinds of experiment clearly complement one another, most experimental studies of thermal adaptation have occurred in the laboratory.

E V O L U T I O N A RY T H E R M A L B I O L O G Y

15

Box 1.3 Measuring selection gradients

where α equals the intercept, β equals the slope of the relationship between the phenotype and ﬁtness, and ε equals the residual variation (i.e., the variation in ﬁtness not associated with the trait); by assumption, ε is normally distributed with a mean of zero. The selection gradient (β) approximates the strength of directional selection (Fig. 1.6a). The selection gradient can be standardized by the mean or the variance of the trait to facilitate comparisons among traits or taxa (Hereford et al. 2004). Stabilizing or disruptive selection is quantiﬁed by adding a second-order term (z 2 ) to the model; a negative β for the second-order term indicates stabilizing selection whereas a positive second-order term indicates disruptive selection (Fig. 1.6b and c). Because most traits are correlated with other traits, researchers often use multiple regression to estimate selection gradients. A multiple regression model describes the relationship between two or more traits and the relative ﬁtness: w = α + β1 z1 + β2 z2 + · · · + βn zn + ε,

(1.11)

where zn and β n equal the phenotypic value and the selection gradient of trait n, respectively. The selection gradient calculated by multiple regression describes the effect of trait n on relative ﬁtness, independent of all other traits (Lande and Arnold 1983). In the context of thermal adaptation, the traits would represent phenotypes measured at different temperatures or parameters of a thermal reaction norm (see Section 1.4.2). In practice, ﬁtness is notoriously difﬁcult to deﬁne, much less measure. Ideally, one would measure the lifetime reproductive success of a sexual organism or the population growth rate of a clonal organism, but many other proxies for ﬁtness have been used to quantify selection (Endler

1.5.3 Comparative analysis Although direct measures of selection are ideal, such observations are impractical or even impossible to accumulate for most species. Given this limitation,

(a)

Relative fitness

(1.10)

␤ (z) > 0 ␤ (z) = 0

b(z) < 0

(b) Relative fitness

w = α + βz + ε,

1986; Kingsolver et al. 2001b). When a proxy for ﬁtness is measured during part of the life cycle, selection gradients describe an episode of selection rather than natural selection, per se (for an insightful discussion of these concepts, see Endler 1986).

␤ (z 2 ) < 0

(c)

Relative fitness

The direction and magnitude of natural selection can be quantiﬁed by regression (reviewed by Brodie et al. 1995). A regression model relates the phenotypic value of a trait (z ) to the relative ﬁtness of an individual (w ; ﬁtness of the individual divided by the mean ﬁtness of the population):

␤ (z 2 ) > 0

Phenotype (z) Figure 1.6 Hypothetical ﬁtness landscapes for (a) directional, (b) stabilizing, and (c) disruptive selection. The sign of the selection gradient (β ) differs among these functions.

the generality of current models must be assessed through comparative analyses. The comparative method has always been the mainstay of biologists (Somero 2000), but we can now choose from a

16

T H E R M A L A D A P TAT I O N

Box 1.4 Experimental designs for detecting adaptation Experiments in the laboratory or ﬁeld enable one to document thermal adaptation, but a careless design can lead to an ambiguous or erroneous result (Rose et al. 1996). An adequate experimental design would enable us to attribute the evolutionary response to natural selection imposed by environmental temperature, as opposed to other selective pressures or genetic drift. In laboratory experiments, one would start with an ancestral population that was already adapted to the artiﬁcial environment. In ﬁeld experiments, populations would be chosen to avoid recent changes in the selective environment. From the ancestral population, one must establish both experimental and control lines (Fig. 1.7). Experimental lines experience a novel thermal environment, whereas control lines continue to experience the thermal conditions of the ancestral population. Because genetic drift within lines can produce spurious patterns or constrain adaptation, each line should consist of a large number of mating pairs (≥50). As in any experiment, adequate replication underlies the quality of the statistical inference. Replicating observations in a selection experiment poses a major challenge because the effective unit of replication equals the population (or line) rather than the individual. Thus, experiments with only a few lines per thermal environment should be interpreted

with caution. That said, most selection experiments presented in this book involved fewer than ﬁve lines per temperature. This unfortunate reality stems partly from a tradeoff imposed by logistical constraints; when the availability of space, time, or money limits the sample size, researchers face a tradeoff between having enough individuals within lines to avoid genetic drift and having enough lines within treatments to achieve statistical power. Inferring thermal adaptation from ﬁeld experiments requires some caution. In an artiﬁcial mesocosm, environmental temperature can be manipulated while all other factors are controlled or randomized among lines. Consequently, one can infer thermal adaptation by comparing experimental lines from a novel thermal treatment with control lines from the ancestral thermal treatment. In a natural environment, many factors covary with environmental temperature. Thus, we might falsely conclude that thermal adaptation occurred if other environmental factors imposed the predicted selective pressure. A more robust inference requires replicated populations in multiple thermal environments, which increases the likelihood that other factors will vary randomly with respect to environmental temperature.

Ancestral population

Control lines

Replicate line 1

Replicate line 2

Experimental lines

Replicate line 1

Replicate line 2

bewildering array of statistical procedures for combining phylogenetic, environmental, and phenotypic data to test evolutionary hypotheses. Understanding these procedures will be helpful because much of the theory presented in this book will be tested by the comparative method. Two applications of the comparative method serve our

Figure 1.7 Replication of experimental evolution requires multiple lines, each founded from an ancestral population. Individuals are randomly sampled from the ancestral population to establish each line.

purpose: (1) studies of covariation between traits to test hypotheses about tradeoffs and (2) studies of covariation between environmental temperature and organismal phenotypes to test hypotheses about thermal adaptation. The simplest comparison involves the study of organisms in extreme environments. Biologists

E V O L U T I O N A RY T H E R M A L B I O L O G Y

17

Box 1.5 Calculating phylogenetically independent contrasts Imagine a trait, X , that has evolved within a lineage represented by four extant species (A–D). The values of X in these species are interdependent because all four species shared some portion of the evolutionary history (Fig. 1.8). Yet each species also experienced a period of evolution that was independent of the other species. For example, species A and B diverged from a common ancestor (E), and any divergence in X that occurred after these species formed was clearly independent of evolution that occurred elsewhere within the lineage. Therefore, the contrast between the values of X in species A and B (i.e., XA − XB ) is called a phylogenetically independent contrast (Felsenstein 1985). For any set of n data points one can compute n − 1 independent contrasts. To compute these contrasts, we need a cladogram depicting only the relationships among extant taxa, branch lengths for the cladogram, and data for two or more variables (Garland et al. 1992). Contrasts are computed for each pair of taxa, starting at the tips and moving down the tree. The values of X for ancestral taxa, such as E in Fig. 1.8, are estimated as the weighted mean of X for its descendants. For example, the value of X for taxon E (XE ) would be calculated as: (1/vA )XA + (1/vB )XB , (1.12) XE = 1/vA + 1/vB where vA and vB equal the branch lengths leading from the ancestor to taxa A and B, respectively. Because pairs of taxa that diverged more recently will have smaller contrasts, one should standardize contrasts before using them in statistical analyses (Garland et al. 1992). Standardization is quite simple because the expected mean of the contrasts equals zero; therefore, all one has to do is divide the contrasts by their standard deviation. We can use contrasts in any statistical analysis, but the following restrictions hold: (1) all standard assumptions of parametric statistics that apply to the raw data also apply to independent contrasts and (2) statistical models should not include an intercept because the intercept of a relationship between two sets of contrasts theoretically equals zero (Garland et al. 1992). The validity of independent contrasts rests on two major assumptions (Garland et al. 2005; Martins and Hansen

automatically assume that organisms living in extreme thermal environments, such as Antarctica, have adaptations for doing so (Magnum and Hochachaka 1998). This assumption seems reasonable given most organisms cannot tolerate such

1996). First, evolution of a trait must have followed a process that resembles Brownian motion; in other words, the expected change in the trait’s value equals zero and the variance of this change scales in proportion to the duration of evolution. This assumption holds for directional selection as long as the optimal phenotype ﬂuctuates over time. Second, the sequence and timings of divergence among taxa are known. Although these assumptions are surely violated in all cases, comparative analyses of raw data require equally restrictive assumptions (Garland et al. 1999). When conducting comparative analyses of thermal adaptation, researchers often calculate independent contrasts for environmental temperature. Arguably, the use of independent contrasts for environmental temperature is inappropriate because environmental variables do not necessarily possess the same properties as phenotypes (Martins 2000). Nevertheless, Garland and colleagues (1992) noted that contrasts could be computed for a continuous environmental variable if environmental states were passed from ancestor to descendant. This assumption seems reasonable considering that geographic structure among populations ensures that individuals in a relatively cold environment produce offspring that experience a relatively cold environment.

XB B

XA A

νA

XC C

XD D

νB E

F

Figure 1.8 This hypothetical cladogram depicts the relationships among four extant taxa, A–D, that differ in some trait X . Ancestral taxa are labeled as E and F. Branch lengths for taxa A and B are labeled as vA and vB , respectively.

thermal extremes. By comparing the phenotypes of these organisms with those of organisms in less extreme environments, we learn how organisms deal with thermal stress. But how extreme must a thermal environment be for us to assume that an

18

T H E R M A L A D A P TAT I O N

organism’s thermal physiology reﬂects adaptation? Because we usually cannot answer this question, comparisons between two taxa from disparate environments provide only weak evidence of thermal adaptation (Garland and Adolph 1994). A more powerful way to study adaptation is to compare phenotypes of species across a range of environments in a phylogenetic context (Gittleman and Luh 1994; Harvey and Purvis 1991; Huey 1987; Martins 2000; Miles and Dunham 1993). Three methods of phylogenetic comparative analysis have been most commonly used (Garland et al. 2005): (1) independent contrasts (Felsenstein 1985), (2) generalized least-squares models (Martins and Hansen 1997), and (3) Monte Carlo simulations (Martins and Garland 1991). Comparative studies of thermal adaptation have used the method of independent contrasts. This method transforms a set of raw data into a set of new data that meet the assumption of statistical independence. These new data consist of contrasts, which estimate the divergence in a trait between two taxa since their most recent common ancestor (Box 1.5). In principle, independent contrasts can be calculated for either phenotypic or environmental variables (Felsenstein 2002; Garland et al. 1992). Therefore, we can use independent contrasts to relate thermal change to evolutionary change within a lineage. In certain circumstances, we should rely on analyses of raw data instead of analyses of independent contrasts. For example, independent contrasts should not be used when phenotypic differences among taxa are caused primarily by environmental factors (i.e., phenotypic plasticity). Likewise, independent contrasts would be undesirable if we were testing the quantitative predictions of competing models, particularly when the models predict nonlinear relationships between environmental temperature and phenotypic values. The reason is simple: independent contrasts do not retain information about the actual environmental or phenotypic values—only the difference between the values for sister taxa (see Box 1.5). The extensive use of independent contrasts stems from the overabundance of qualitative predictions in evolutionary

biology (i.e., trait X should correlate positively with trait Y). Finally, an analysis of independent contrasts would be an overly conservative test of adaptation if stabilizing selection has maintained the observed phenotypes (Martins 2000). For all of these reasons, we shall interpret analyses of raw data in conjunction with analyses of independent contrasts when the two disagree.

1.6 Conclusions As an opening to this book, I have tried to lay a foundation on which we can rest the concepts, models, and experiments presented in the coming chapters. The array of phenomena studied by thermal biologists—thermal sensitivity, thermoregulation, and thermal acclimation—naturally lead to the construct of a thermal reaction norm. This construct opens many doors because biologists have spent decades developing tools for the study of reaction norms. By estimating and comparing thermal reaction norms for labile or ﬁxed traits, we can discover general patterns of interest. But to understand the evolution of these patterns, we must do more than accumulate data. We must gather and interpret our data in light of models. Evolutionary biology offers the theoretical and empirical tools needed to dissect these patterns and identify their causes. The evolution of thermal reaction norms can be explored through optimality models, quantitative genetic models, and allelic models. As we shall see, these models offer a wealth of predictions about thermal adaptation. The accuracy and generality of these models can be determined from direct measures of natural selection, experimental evolution in controlled environments, and comparative analyses of phenotypic variation. Before we begin to use these tools, we must familiarize ourselves with the thermal heterogeneity of natural environments. This variation in temperature over space and time creates the impetus for evolutionary change. For this reason, we shall deal with the topic of thermal heterogeneity in the next chapter.

CHAPTER 2

Thermal Heterogeneity

The temperature of an organism—a quantitative measure of the kinetic energy of its molecules— determines the capacity for heat to ﬂow between the organism and its environment (see Haynie 2001). If the organism possesses the same temperature as its environment, the two are in thermal equilibrium (i.e., no net ﬂow of heat occurs between them). But if the organism and the environment differ in temperature, the warmer entity will lose heat while the cooler one gains it. This ﬂow of heat occurs through a variety of physical processes whose relative importance varies among environments (Gates 1980; Mitchell 1976; Porter and Gates 1969). Aquatic organisms mainly exchange heat with their environment via conduction and convection. The high thermal conductivity and heat capacity of water cause the body temperatures of these organisms to closely follow the temperatures of their aqueous surroundings (Spotila et al. 1992). For terrestrial organisms, conduction, convection, radiation, and evaporation contribute signiﬁcantly to the body temperature (Fig. 2.1). Below the surface of the ground, the impacts of radiation, convection, and evaporation weaken (as they do in water). These processes cause a net exchange of heat between an organism and its environment as long as the two differ in temperature. In the real world, organisms rarely enter thermal equilibria with their environments. The motion of the earth combined with radiation from the sun drives a continuous redistribution of heat throughout the planet. Thus, living things must deal with thermal change on a variety of temporal scales. First, environmental temperatures cycle daily because of the periodic exposure to solar radiation; this predictable source of thermal heterogeneity results from the rotation of the earth around its axis. Second,

environmental temperatures change rapidly and unpredictably with atmospheric conditions, such as wind speed and cloud cover. Third, environmental temperatures change seasonally because of the tilt of the earth as it orbits the sun. Finally, environmental temperatures change both cyclically and directionally among years. For example, changes in atmospheric and oceanic circulation drive fairly regular phenomena of warming and cooling in the Paciﬁc Ocean, referred to as El Niño and La Niña. In geological history, extensive periods of warming and cooling were triggered by changes in the earth’s orientation and orbit. In recent history, humans have perturbed the atmosphere so much that earth has entered a period of warming comparable to those caused by astronomical events. Depending on its location and lifespan, an organism experiences the combined stress of some or all of this thermal change. In this chapter, we shall consider the patterns of thermal heterogeneity experienced by organisms. As we shall see, these patterns can be quantiﬁed with mathematical, physical, or statistical models. In each case, one seeks to deﬁne the body temperature of an organism based on variables that inﬂuence heat exchange. First, we shall derive a mathematical relationship between environmental conditions and the body temperature of an organism. Then, we shall use this relationship to explore how spatial and temporal variations in environmental conditions create heterogeneity of body temperatures on a global scale. Finally, we shall see how physical and statistical models have enabled researchers to quantify thermal heterogeneity on local scales. The patterns of environmental variation described in this chapter will help us to deﬁne the selective pressures on physiology, behavior, and life history in subsequent chapters.

19

20

T H E R M A L A D A P TAT I O N

Dust

Radiation from the atmosphere

Scattered solar radiation Reflected solar radiation Evaporative heat loss Convective heat exchange

Radiation from vegetation

Radiative heat loss

Radiation from the ground

Conductive heat exchange

2.1 Operative environmental temperature When an organism’s environment changes, so does its body temperature. Consider a terrestrial organism consisting of a body core and a body surface. This individual gains heat from its metabolic reactions and the radiation absorbed by the body surface. Incoming radiation includes short-wave radiation from the sun (solar radiation) and long-wave radiation from all other surroundings (thermal radiation). The individual loses heat through the emission of radiation from the body surface and the evaporation of water from the respiratory organs and body surface. Heat is also exchanged with the ground and air by conduction and convection, respectively. The transfer of heat by these processes occurs (1) between the core and surface of the organism and (2) between the surface of the organism and its environment (Fig. 2.2). When environmental conditions change, these ﬂuxes of heat change. The organism will heat or cool until it reaches a steadystate temperature. At this steady state, the organism continues to exchange heat with its environment, but gains and losses balance one another.

Figure 2.1 Routes of heat exchange between an organism and its environment include radiation, conduction, convection, and evaporation. An organism absorbs several kinds of radiation, including direct radiation from the sun, scattered radiation from the atmosphere, and reﬂected and thermal radiation from surfaces. Adapted from Tracy (1976) with permission from the Ecological Society of America.

Qmet – Qresp

Qevap

Qabs

Kcond Qemit

Ts

Tb

Kconv Ta Kcond Tg

Figure 2.2 A thermal circuit diagram depicts the routes of heat transfer between a terrestrial organism and its environment. Nodes representing the body core, body surface, air, and ground are labeled according to their temperatures (Tb , Ts , Ta , and Tg , respectively). A resistor symbol denotes heat ﬂux by conduction or convection, whereas an arrow denotes any other form of heat ﬂux. Rates of conduction and convection depend on the coefﬁcients of conduction (Kcond ) and convection (Kconv ), respectively. The capacitance symbol represents the heat capacity of the animal. Adapted from Bakken (1976) with permission from Elsevier. © 1976 Elsevier Ltd.

To understand how certain environmental conditions inﬂuence the body temperature of an organism, we shall apply a concept known as the operative environmental temperature (or operative temperature). The operative temperature (Te ) equals the

THERMAL HETEROGENEITY

steady-state temperature of an organism in a particular microclimate in the absence of metabolic heating and evaporative cooling (Bakken 1992). As such, this temperature characterizes the thermal environment as perceived by the organism, independent of any physiological thermoregulation. To deﬁne the operative temperature in mathematical terms, Bakken and Gates (1975) modeled the routes of heat ﬂux between an organism and its environment. Gains and losses at the surface of the organism were described as follows: Qcond + Qabs = Qevap + Qemit + Qconv ,

(2.1)

where Qcond , Qabs , Qevap , Qemit , and Qconv equal rates of heat ﬂux caused by conduction, absorbed radiation, evaporation, emitted radiation, and convection, respectively. The term for conduction (Qcond ) refers to heat ﬂux between the core and surface of the body; conduction between the body surface and the environment was assumed negligible.4 Similar ﬂuxes determine the change in the organism’s core temperature over time: C

dTb = Qmet − Qresp − Qcond , dt

(2.2)

where C equals the heat capacity of the organism, Tb equals the core body temperature, t equals time, and Qresp equals the rate of evaporative heat loss during respiration. These rates of heat ﬂux represent averages for the whole organism, which avoids an explicit consideration of spatial heterogeneity or surface area. In reality, these ﬂuxes vary among regions of the organism, but this simpliﬁed model will sufﬁce for our purposes (for a more complex model, see Bakken 1981). At a steady state, the inﬂux and efﬂux of heat balance one another such that the core and surface temperatures of the organism remain constant over time. To ﬁnd the steady state, we must ﬁnd the core temperature that satisﬁes the conditions set by

21

eqns 2.1 and 2.2, where dTb /dt = 0. Solving for the steady-state temperature becomes much easier when the terms in these equations are linear with respect to temperature. For example, conduction and convection are linear functions of differences between temperatures: Qcond = Kcond (Tb − Ts )

(2.3)

Qconv = Kconv (Ts − Ta ),

(2.4)

and

where Kcond and Kconv equal coefﬁcients of conduction and convection (W K−1 ), and Ts and Ta equal temperatures of the body surface and air, respectively. In contrast, emitted radiation scales nonlinearly with the temperature of the organism’s surface: Qemit = Aσ εTs4 ,

(2.5)

where A equals the surface area of the organism, σ equals the Boltzmann constant, and ε equals the emissivity of the organism. To simplify calculations of heat transfer, Bakken (1976) approximated this nonlinear term with a linear one:5 Qemit ∼ = Aσ εTa4 + 4Aσ εTa3 (Ts − Ta ).

(2.6)

With this approximation, eqn. 2.1 becomes: Kcond (Tb − Ts ) + Qabs ∼ = Qevap + (Aσ εTa4 + 4Aσ εTa3 (Ts − Ta )) + Kconv (Ts − Ta ).

(2.7)

The approximation of emitted radiation causes little error, except when the body temperature of the organism deviates substantially from the air temperature (Tracy et al. 1984). Once we have linearized eqn 2.1, we can easily solve for the steady-state temperature of the organism. To do so, we ﬁrst substitute eqn 2.3 into

4

One can reasonably ignore conduction between the organism and its environment when little surface area of the organism contacts the ground or when little difference exists between the temperatures of the ground and the body surface. Bakken (1976) presented a more general model of operative temperature that includes conduction between the organism’s core and the ground, but subsequent applications of the concept of operative temperature follow the special case derived here (see, for example, Bakken 1985; Bakken 1992; Dzialowski 2005).

5

Bakken and Gates (1975) used a different linear function to approximate radiation, resulting in a slightly different mathematical deﬁnition of operative temperature. Here, I combined the linear approximation of Bakken (1976) with the model of heat exchange proposed by Bakken and Gates (1975). This approach leads to the deﬁnition of operative temperature that appears in subsequent publications (G.S. Bakken, personal communication).

22

T H E R M A L A D A P TAT I O N

eqn 2.2 and rearrange to deﬁne Ts in terms of Tb : Ts = Tb −

b Qmet − Qresp − C dT dt

Kcond

.

(2.8)

Then, we obtain the steady-state temperature of the organism by substituting eqn 2.8 into eqn 2.7, setting dTb /dt equal to zero, and solving for Tb : Tb ∼ = Ta + +

Qmet − Qresp Kcond

Qmet + Qabs − Qresp − Qevap − Aσ εTa4 Kconv + 4Aσ εTa3

.

(2.9) If we assume the organism gains no heat from metabolism and loses no heat by evaporation (or that these ﬂuxes cancel one another), we arrive at the operative temperature: Qabs − Aσ εTa4 Te ∼ . = Ta + Kconv + 4Aσ εTa3

(2.10)

Thus, the operative temperature reﬂects an increment in body temperature caused by solar radiation, discounted for the heat lost by convection. In a burrow, where an organism receives no solar radiation, the operative temperature equals the ambient temperature (Chappell and Whitman 1990). For certain organisms, the operative temperature should account for the heat lost by evaporation. For example, amphibians possess a wet epidermis and many species cannot physiologically control their evaporative water loss (Duellman and Trueb

1986; Spotila et al. 1992). The relative humidity of the environment determines the rate of evaporation, and hence inﬂuences the body temperature of these organisms (Gates 1980). A biologically relevant measure of operative temperature for such organisms, referred to as the wet operative temperature, accounts for evaporative heat loss (for details, see Campbell 1977). Operative temperatures depend on interactions among absorbed radiation (Qabs ), air temperature (Ta ), and wind speed (which affects Kconv ). We can visualize these interactions in the form of a climate–space diagram (Kingsolver 1983; Porter and Gates 1969; Scott et al. 1982; Spotila et al. 1972). These diagrams depict the combinations of environmental conditions for which an organism will experience a certain steady-state temperature. Obviously, these combinations depend on the physical properties of the organism that determine heat exchange (e.g., color, shape, and size). Figure 2.3 shows the combinations of absorbed radiation, air temperature, and wind speed for which a desert iguana (Dipsosaurus dorsalis) will reach 30◦ C. The impact of absorbed radiation and air temperature on the steady-state temperature depends on the wind speed, as depicted by the crossing of the lines. Such plots highlight the interactions among environmental factors that inﬂuence the operative temperature. To the extent that these factors vary predictably over space and time, they create patterns of thermal heterogeneity.

40 Wind = 0 cm s–1 Wind = 100 cm s–1 Wind = 1000 cm s–1

Air temperature (°C)

35 30 25 20 15 10

0.5

0.6

0.7

0.8

0.9

Absorbed radiation (cal cm–2 min–1)

1.0

Figure 2.3 The same body temperature occurs under many combinations of radiation load, air temperature, and wind speed. Each line shows the conditions that yield a predicted body temperature of 30◦ C for a desert iguana, with a mass of 67 g and a surface area of 180 cm2 . Adapted from Porter and Gates (1969) with permission from the Ecological Society of America.

THERMAL HETEROGENEITY

2.2 Global patterns of operative temperature Global variation in solar radiation and environmental temperature gives rise to latitudinal and altitudinal clines in operative temperatures. The intensity of solar radiation determines the amount of radiation absorbed by a particular organism. The air or water temperature determines the heat exchanged via convection. As solar radiation and environmental temperatures vary dielly and seasonally, so do geographic clines in operative temperatures.

2.2.1 Latitudinal clines Ultimately, all latitudinal variation in operative temperatures stems from changes in solar radiation. The quantity of solar radiation received by an organism and its environment depends on the time of day, the time of year, and the location on earth (reviewed by Gates 1980; McCullough and Porter 1971). These sources of variation stem from the tilt of the earth’s axis and its elliptical orbit around the sun. During June, the northern hemisphere tilts toward the sun, and thus receives more solar radiation that the southern hemisphere. The opposite phenomenon occurs during December, leading to the opposing seasons observed in these regions of the planet. At any location, the rotation of the earth on its axis creates variation in solar radiation on a daily basis; radiation peaks daily at solar noon, when the sun’s rays become perpendicular to a particular surface of the earth. These simple principles combine to determine the radiation received at a particular place and time, h S0 (Gates 1980): 2 d¯ ¯ (sin φ · sin δ + cos φ · cos δ · cos h), h S0 = S 0 d (2.11) where S¯ 0 equals the solar constant (1360 W m−2 ), d and d¯ equal the actual and mean distances respectively between the earth and the sun, φ equals the latitude of the surface, δ equals the declination of the sun, and h equals the hour angle of the sun (the angle by which the earth must rotate to bring the point directly perpendicular to the sun’s rays). Given the presence of an atmosphere, only a fraction of solar radiation reaches the surface of the

23

earth; the atmosphere absorbs, reﬂects, or scatters the remaining radiation. To model the reduction in solar radiation caused by the atmosphere, we must add a coefﬁcient (τ m ) that deﬁnes the transmission of radiation: 2 d¯ ¯ S = S 0 h 0 d × (sin φ · sin δ + cos φ · cos δ · cos h)τ m , (2.12) where 0 ≤ τ ≤ 1. The exponent m equals the length of the path by which radiation travels through the atmosphere divided by the shortest possible path to sea level. By integrating eqn 2.2 over time, one can calculate daily or annual radiation load for any latitude. The hour angle changes over the course of a day, and the sun’s declination changes over the course of a year. Gates (1980) integrated over the set of hour angles to obtain the daily radiation load; plotting this radiation as a function of latitude and time of year reveals certain patterns (Fig. 2.4). For a realistic range of atmospheric transmissions, the greatest incident radiation occurs at 30◦ N during the solstice in June and 30◦ S during the solstice in December (Fig. 2.5a). During the equinoxes in March and September, however, the greatest incident radiation occurs at the equator (Fig. 2.5b). Over the course of a year, the equator receives 2.4 times the radiation that the poles receive (Gates 1980). This variation in solar radiation contributes directly to latitudinal clines in the operative temperatures of organisms (see eqn 2.10). Solar radiation also affects operative temperatures indirectly by causing variation in the temperatures of air, water, and soil. Global patterns of environmental temperature are most easily inferred from air temperatures, which have been recorded for decades at thousands of locations (Box 2.1). In the northern hemisphere, air temperature decreases strongly with increasing latitude (Fig. 2.7). This latitudinal gradient becomes most pronounced during winter (Addo-Bediako et al. 2000; Bradshaw and Holzapfel 2006; Ghalambor et al. 2006), a phenomenon resulting from the contrasting seasonal patterns of solar radiation at different latitudes (see Fig. 2.4). A global model of daily temperatures constructed by Piper and Stewart (1996) nicely

24

T H E R M A L A D A P TAT I O N

90 0

0

North

60 10

10

Latitude

30

0 10

South

–30

–60

0

Ja n. Fe b. M ar . A pr . M ay Ju n. Ju l. A ug . Se p. O ct . N ov . D ec .

–90

Month

Figure 2.4 The daily intensity of incident radiation varies predictably with latitude and the time of the year. The values plotted here were based on a transmittance of 0.6. Adapted from Gates (1980) with permission from Dover Press.

(a) 50

(b) 40

= 1.0

= 1.0

Radiation (MJ m–2 day–1)

Radiation (MJ m–2 day–1)

40

= 0.8

30

= 0.6 20

= 0.8

20

= 0.6

10

10

–90

30

–60

–30

0 30 Latitude

60

90

–90

–60

–30

0 Latitude

30

60

90

Figure 2.5 The transmission of solar radiation affects the latitudinal clines of solar radiation during the solstice (a) and the equinox (b). Positive values of latitude are for the northern hemisphere. Adapted from Gates (1980) with permission from Dover Press.

THERMAL HETEROGENEITY

Box 2.1 Mapping environmental temperatures on a global scale Since the production of the ﬁrst standardized thermometers in the 18th century, the measurement and analysis of temperatures have blossomed into a global enterprise. Today, the World Meteorological Organization, the World Data Center for Meteorology, and the Intergovernmental Oceanographic Commission collectively archive temperatures recorded by roughly 10 000 terrestrial weather stations and several hundred coastal buoys. Additionally, thermal records for the ocean surface have been estimated from satellite images (Kerr and Ostrovsky 2003). To facilitate comparative research, these organizations set guidelines for methodology and accuracy; for example, the World Meteorological Organization deﬁnes air temperature as the temperature of a thermometer exposed to air at a height of 1.2–2.0 m and sheltered from direct solar radiation. The efforts of these organizations have generated two beneﬁcial outcomes for thermal biologists. First, we now have excellent models of spatial and temporal variations in air and water temperatures to guide our research. Second, we can use normal temperatures—mean temperatures for a consecutive period of 30 years—to identify anomalous conditions and directional trends. Current estimates of global warming depend on normal temperatures for 1961–1990 (see Chapter 9). The thermal data collected by weather stations enable one to map environmental temperatures on a global scale. To construct such maps, one must convert the extremely patchy distribution of thermal records into a regularly spaced grid. Climatologists accomplish this feat by ﬁtting regression models to the available data or interpolating temperatures between weather stations (see, for example, Zhao et al. 2005). The global patterns described in this chapter stem from the work of Piper and Stewart (1996) and New and colleagues (1999), who used methods of interpolation to map mean temperatures and mean daily ranges of temperatures for all continents but Antarctica. Below, I brieﬂy contrast the procedures used by these investigators. Piper and Stewart (1996) used time series of air temperatures from more than 7500 weather stations to create maps of daily mean temperatures and daily ranges of temperature at a spatial resolution of 1◦ (latitude by longitude). To increase the size of their sample while ensuring consistency among stations, they recalculated mean temperatures as the average of the minimal and maximal temperatures. To simplify their model, they adjusted all temperatures to sea level prior to interpolation.

After interpolation, they readjusted the temperature at each location based on the altitude. Their method of interpolation involved distance weighting. For each grid point, the temperature was estimated from temperatures recorded at 5–10 neighboring weather stations; the inﬂuence of each record on the predicted temperature was weighted by the proximity of the weather station to the grid point. In contrast to Piper and Stewart, New and his colleagues used a two-step process to generate maps of monthly mean temperatures and monthly ranges of temperatures at a resolution of half a degree latitude/longitude. First, they modeled the global distribution of normal temperatures for 1961–1990 (New et al. 1999). Because normal temperatures are more accessible than time series, the researchers could include data from more than 12 000 locations. Each temperature was weighted by the variance and the number of years contributing to the normal temperature (maximum = 30 years). Unlike Piper and Stewart, New et al. interpolated over three dimensions, eliminating the need to adjust temperatures for the altitude. To avoid errors caused by the low density of weather stations in some regions, they divided the land surfaces into 22 overlapping tiles (Fig. 2.6) and ﬁtted a separate spline function to the data for each tile. Once New et al. had obtained a grid of normal temperatures, they used records of thermal anomalies (deviations from the normal temperature) to create monthly thermal maps for the period of 1901–1996 (New et al. 2000). Thermal anomalies were interpolated by distance weighting to form a grid for each month. By combining the grid of thermal anomalies with the grid of normal temperatures, New et al. generated a global map of temperatures for each period of time. Several sources of error inﬂuenced the quality of global thermal maps. First, not all measurements of temperature were made at comparable times of day (Piper and Stewart 1996). Deriving daily mean temperatures from minimal and maximal temperatures reduced some bias caused by differences in timing (New et al. 1999). Second, the sheer magnitude of data involved in these analyses virtually guaranteed the existence of typographical errors. Many of these errors were eliminated by careful review of the data prior to analysis. Errors were assumed to exist when monthly mean temperatures failed to follow a seasonal cycle or a mean temperature fell outside the range bounded by the minimal and maximal temperatures. Finally, errors in interpolation might have resulted from a paucity of temperatures at certain latitudes and altitudes. continues

25

26

T H E R M A L A D A P TAT I O N

Box 2.1 Continued 80

. . .. . . . . ..... . . . . 16 . .. 10 . . . . . . . . . . . . . .. . . . . . . . . .. . . . . .. . .... . . . . . . . . .. . . . . . . .. . . . . . . . . . . . . . . . . .. . .. .................. ... . .... ....... . . . . . . .. . . . . .. . .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . ... .. ... ... .. . . . .. . . . . .. . .. . . . .. .. . .. ... .. ....... . ... .... . . . .. ................. . .. . . ..... . ....... ... ................... . ... . . .. ... ... . .. . . . . . .. . .. .. . . .. . . .. .. . .. . . ............ ...... ..... .. ........ ........ . .. .. . .. ............................ ......................... ........... .......... .. . . ......... . . .. .. ... ........ . . .. . . . . . .. . .. . . . . . . . . .. . . . . . .. .. ...................... .................... .. . . .. . . . . . . . .. . . . . . .. . ... . . . . . . . . . . . . . .. .... . . . . .. . . . . . . . . . . . . . . . . . .. . . . . . . . .. . .. . . . . . . . . . . . . . . . .. .. . . . . . . . . . .. .. .. . . . . . . . . ... ... . ... . . . . . .. . . .............. .. 2 .................. ........... . .. . .. . . .. .. .. . . ......... ... . ........... .. .................. .. . .............. ..... ... ................................ ...................... ............... ...... ......................... . .... .......... ... ... ................ .......... .. ........................ ... . ....... .. ...... ............................ . ....... . .................... ........................ ... ........................ .... .................... .... .. .. . . .... . ..... .... ...................... ..... ..................................... ................ ................................... ............................. . ... .. .. . .. . . . ..... ............ . .... .... ...... ...... ... .. ......... ................................. ........ .. .............. .............. ... ....... ..... .. ... ..... ............... ..... ...... .. ...... .................... ........ ...... ....... ............................. ... .... ..... .. ........... 3 ...................................................................... .. ...................................... ... ............. . .......... ...... ..... . .. ...... ..... ........ . ....... .. .............. ..... ... 11 ........ ...... ............ ...... ..... . .................. ........... ...... ........... .......... ... .... ........................ ........... ....... ....... .......... ................. ........................ .............. .... ............................. . ................ ..... ......... . . . . . ...... . . . . . ... . .. . . . . . . . . . . ... . .. . ..... . . .. .................. ....... . . . . . . . . . . .. .. . . . . . .. ... . . . . . . . . . . .. . . . ... . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .. . .. . . .. . . . .. . .. .. .. . . .. .. . ... .. ..... ........... .............. .............. ............. .... ......................... ................................. .......... ..... ... .... .. ..... ........... ........... .... ... .............. ........................ ................ . ........ ................... .. ..... ........ ....... . . ...... ............... .... ................ . ..... ....... ....... ............... ................ ... ............. .. ....... .................. . .. ......... ...................................... ....... ............................ ... ... ...... . ... .. .......... ............ ......... . . ......... 4 ................................ .. ... ..... ....................... ........ ...... .... .. . ...... ......... ... ........ . ..... ... .. ... .. ............... ......... .......... ......... ....... ....... . .... ...... ............... ........ 19 ....................... . .. ..................................................... ................. .................. .... . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . .. . . . . . . ...... . . . . . .. . . . .. . . . . . . .... . .. . . . . ... . . . . . . . . . . .. . . . . ... ... ..... . ...... ..... .. . ..... . ..... . .... . ..... 5 . ....................................................................... ..... .................................. . .. . . ..... . ...... ...... ................. ............. ........... . . .... . .. . ..... ......... ...... 18 .......................... . .. . . . . . .. . . . . . . . . . . . . .. . .. .. . . . .. . .. . . . . . ......... ........... .... .... ... . . .. . . ............. .. ........ ......... .. .......... ............... .. .... ... . ... . 6 . . . . .. .. .. .. . . .. . . . .. . . .. . . . . . . . . . . . ......... ... ... ... ........... .............. . ... ............... ... ....... . . . . . ............. .. . . .............. .. . . . . . . . . . . . ................................................. . . . . . . .. . . 17 . . . .. .. .......... ........................... ... ... ..... ........ .............. . . ...... . . . .. 7.. ........................... ........ .. .. .... ... ......... . ........13 ............................... ................... ........ 12 .. ....... ... . . . . ... . . ....... .. .. . . . .... ..... . . . ..... ........ . . .......... . . .. . ............... ....... . . . ... . . . . . . . .. . . . . . . . ... . . . . ..... .. . ........... .. .. . . . .......... . . . ... . .. ..... .................... . . ..... . . . .. ..... . . ...... ........ . .. .. . 20. ..................... . . . ......... .... .. . ... . . .. . ..... ... .............. . ................. .. . .... ..... ......... .. ..... . ................. . . . . . ...... ..... ... . ... ... .. .. ...................... . ..... ......... . .. . ... . . . . . . . .. ..... ......... . .......... . 8 . . . . . . . . . . . .. ... . . . . . . . 14 . . .. .. . . . . . . . . . . . . . . . . . . . . . . .. . . ..... ..... . . . . . ..... ....................... .. . .. ... . . ......... ....... ........................... .... .. ......................................................... . . . . . . . . . . .. . . . . . . . . . . . .. . . .. . . . . .. . . .. . . . . . . .. . . . . .. . . . . . . . . . .. . . . ... . . .. . ... . . ............................ .. .. ....... ......... . ... ..... ......... . . ... ..... . .... ............ ......... ......... ............................................ . .......... .. ....... 15 . . .............. . ........................................... . ..... ..... ... ................................ . ............. ..... . .... ...... ......... . ... . . . .. ............ . . . 21 ........ . .. . .. .. . . . . . . . . .. . . . . . . . . . ..... . . . 22 . 9 . .

1

..

60 40 20 0 –20 –40

–60 –180

–150

–120

–90

–60

–30

0

30

60

90

120

150

180

Figure 2.6 Locations map of air temperatures. To increase the accuracy of interpolation, the continents were divided into 22 tiles based on the density of weather stations. Reproduced from New et al. (1999) with permission from the American Meteorological Society.

New et al.’s method of ﬁtting of splines to separate tiles helped to minimize the inﬂuence of data-rich regions on data-poor regions. Cross-validation of global thermal

maps indicated these potential sources of error were not so problematic as to obscure major patterns of thermal heterogeneity (New et al. 2000).

30

40

30

10 20 0 10 –10 0

–20

–30 90 75 60 45 30 15

0 –15 –30 –45 –60 –75 –90

Latitude (°)

Temperature range (°C)

Mean temperature (°C)

20

Figure 2.7 Air temperatures of terrestrial environments depend strongly on latitude. Data are the mean air temperatures (black circles) and annual ranges of air temperatures (gray circles) in the New World. Positive values of latitude are for the northern hemisphere. The maximal mean temperature and the minimal thermal range occurs in the tropics. Adapted from Clarke and Gaston (2006) with permission from the Royal Society.

THERMAL HETEROGENEITY

thermal environments (Vázquez and Stevens 2004). In contrast, the southern hemisphere has a weaker latitudinal cline in temperature (Fig. 2.7), owing to the buffering of the small land masses in this hemisphere by vast oceans (Addo-Bediako et al. 2000; Ghalambor et al. 2006). Aquatic environments exhibit far less variation in temperature because of water’s tremendous capacity to store heat. This phenomenon holds for freshwater and marine environments, but is especially exaggerated in large bodies of water such as oceans. For example, the annual range of surface temperatures of the Paciﬁc Ocean never reaches one-half of the annual range observed at the same latitude on land (compare Figs 2.7 and 2.9). This thermal buffering keeps coastal climates warmer than inland ones, particularly in winter (New et al. 2002). Interestingly, while seasonal variation in temperature peaks at high latitudes in terrestrial environments, the peak occurs at intermediate latitudes in the oceans (see Fig. 2.9 and Parmesan et al. 2005). Nevertheless, mean water temperatures of marine and freshwater environments decrease from the tropics to the poles, in a manner that resembles latitudinal clines in air temperature (Clarke and Gaston 2006; Vannote and Sweeney 1980).

40 20 0 –20

Norway (70.5°N, 24.5°E)

–40 40 20 Minimal/maximal air temperature (°C)

0 –20 Belarus (53.5°N, 24.5°E)

–40 40 20 0 –20

Romania (45.5°N, 24.5°E)

–40 40 20 0 –20

Sudan (15.5°N, 24.5°E)

–40

27

40 20 0

2.2.2 Altitudinal clines

–20 Zaire (0.5°N, 24.5°E)

–40 0

60

120

240 180 Day of year

300

360

Figure 2.8 Seasonal proﬁles of daily minimal and maximal temperatures at ﬁve latitudes, ranging from the equator to the Arctic. The longitude was held constant at 24.5◦ E. Modiﬁed from Piper and Stewart (1996) by permission of the American Geophysical Union.

illustrates this point (Fig. 2.8). At the equatorial latitude of Zaire (0.5◦ N), temperatures remain fairly constant throughout the year. Moving from Zaire toward Belarus (53.5◦ N), the seasonal variation in daily temperatures becomes greater. As one moves even farther northward (70.5◦ N), the seasonal range of temperatures decreases slightly because summer temperatures never approach those seen at lower latitudes. The data from these localities reﬂect a very general latitudinal cline in the seasonality of

At any latitude, a shift in altitude results in predictable changes in solar radiation and air temperature. Both of these changes relate to the relatively low air pressure at a high altitude (air pressure reﬂects the mass of air above a point in the atmosphere). Recall that the atmosphere absorbs, reﬂects, or scatters radiation. If all else is equal, a shorter path through the atmosphere results in a greater transmission of radiation. Speciﬁcally, the exponent of the transmission coefﬁcient (m of eqn 2.12) decreases linearly with decreasing air pressure (Gates 1980). Therefore, surfaces at a higher altitude receive a greater intensity of solar radiation. Additionally, the low air pressure at a high altitude reduces the air temperature through an adiabatic process (Monteith and Unsworth 1990). As the pressure of rising air decreases it expands and cools. For dry air, the magnitude of adiabatic cooling equals about 10◦ C per kilometer. Moist air cools less as it rises because of

28

T H E R M A L A D A P TAT I O N

35

12 10

25

8

20 6 15 4 10 2

5

Range of temperatures (°C)

Mean temperature (°C)

30

Figure 2.9 Surface temperatures of marine environments depend strongly on latitude. Data are the mean surface temperatures (black circles) and annual ranges of surface temperatures (gray circles) in the Paciﬁc Ocean (170◦ W). Positive values of latitude are for the northern hemisphere. Latitudinal patterns of ocean temperatures match those of terrestrial air temperatures, except the thermal range of ocean temperatures peaks in each temperate zone and drops sharply toward each polar region. Adapted from Clarke and Gaston (2006) with permission from the Royal Society.

0

0

90 75 60 45 30

(a)

15 0 –15 –30 –45 –60 –75 –90 Latitude (°)

(b) 10

20

8 Number of sites

Temperature (°C)

0

−20

−40

6

4

2

0

−60 0

2

4 6 Altitude (km)

8

10

−8

−7 −6 −5 −4 −3 −2 Temperature lapse rate (°C km–1)

Figure 2.10 Air temperatures decrease with increasing altitude because of adiabatic cooling. (a) Air temperatures at different altitudes measured by a weather balloon released in Germany. Black and gray circles denote measurements taken during summer and winter, respectively. The slope of each line describes the rate of adiabatic cooling. (b) Most rates of adiabatic cooling throughout the globe range from –5◦ C to –7◦ C per kilometer of altitude. Hashed and gray bars denote measurements taken during summer and winter, respectively. Adapted from Dillon et al. (2006) by permission from Oxford University Press.

THERMAL HETEROGENEITY

the heat generated by condensation. We express the consequence of this cooling in the form of an environmental lapse rate, which equals the change in the temperature of still air with an increase in altitude. The environmental lapse rate varies among localities, but usually falls in the range of 5–7◦ C per kilometer (Fig. 2.10). Because of adiabatic cooling, an organism at a high altitude in the tropics can experience an air temperature similar to that experienced by an organism at a low altitude in the temperate zone (Ghalambor et al. 2006). Note, the greater solar radiation and the lower air temperature at higher altitude have opposing effects on the operative temperature of an organism (see eqn 2.10). In general, the effect of air temperature outweighs that of solar radiation, leading to lower operative temperatures at higher altitudes (Adolph 1990; Bashey and Dunham 1997; Beaupre 1995; Diaz 1997; Grant and Dunham 1990; Hertz 1992; Van Damme et al. 1989).

2.3 Quantifying local variation in operative temperatures Because operative temperature depends on absorbed radiation and environmental temperature, environments at higher latitudes or higher altitudes would offer lower operative temperatures if all other factors were equal. Yet all else is never equal! Regional differences in topographical and meteorological conditions disrupt the latitudinal and altitudinal clines in operative temperature caused by solar radiation and environmental temperature. For example, a uniform warming of air by 1◦ C along a latitudinal cline was predicted to increase body temperatures of intertidal mussels by anywhere from 0.1◦ C to 0.9◦ C, depending on the timing and duration of emersion at each site (Gilman et al. 2006). Similar irregularities in geographic clines result from variations in wind speed, cloud cover, and vegetation density. These factors inﬂuence conduction, convection, or radiation to create heterogeneity of operative temperatures on local scales. To really understand the selective pressures imposed by thermal environments, we must model the local variation in operative temperature. Quantifying local variation in operative temperature poses a major challenge. As we can see from eqn 2.10, the operative temperature depends on interactions among micrometeorological conditions

29

of the environment and physical properties of the organism (see also Gates 1980). Micrometeorological conditions vary greatly over very small scales (see, for example, Bartlett and Gates 1967); therefore, we must have some way of directly or indirectly monitoring these conditions at the numerous microenvironments available to an organism. Furthermore, color, shape, and size vary considerably among organisms. Consequently, models of operative temperature must be tailored for each species, or even for each phenotype of a species. Because of these difﬁculties, we know less about the variation in operative temperatures within local environments. What we do know about local heterogeneity comes from the use of mathematical, physical, and statistical models. These models differ in their attention to mechanism and their ease of application. At one extreme, a mathematical model provides a detailed description of heat transfer, but requires careful measurements of many parameters at each location. At the other extreme, a statistical model creates a black box relating environmental characteristics and operative temperatures. Although these models lack an explicit mechanistic link between variables, they offer the potential to predict operative temperatures over much broader scales. At the interface of these two approaches lies the popular approach of measuring operative temperatures with a physical model of the organism. In this section, we shall examine the advantages and disadvantages of each approach.

2.3.1 Mathematical models The earliest attempts to characterize operative thermal environments relied on mathematical models of heat transfer. As we saw in Section 2.1, we can use the principles of biophysics to derive a mathematical function relating operative temperature to micrometeorological variables and organismal parameters. Researchers have developed similar models for a wide range of organisms, including mollusks (Denny and Harley 2006; Helmuth 1998), insects (Henwood 1975; Kingsolver 1983), amphibians (Tracy 1976), and reptiles (Porter et al. 1973; Seebacher and Grigg 1997). Mathematical models have been used to characterize operative temperatures on both regional and local scales.

T H E R M A L A D A P TAT I O N

800

750

0

500 1000 Time (s)

1500 38

500 1000 Time (s)

500 1000 Time (s)

1500

500 1000 Time (s)

1500

35 34 33 32 31 0

500

1000

1500 44

Time (s)

0

0

36

Ground temperature (°C)

32 31 30 29 28 27 26 25

6 5 4 3 2 1 0

Predicted Actual

37

30 Air temperature (°C)

Wind speed (m s–1)

850

Body temperature (°C)

Solar flux (W m–2)

30

1500

43 42 41 40 39

0

Figure 2.11 Measurements of radiation load, air temperature, ground temperature, and wind speed were used to compute the temperature of a mussel on a rocky shoreline. On average, the predicted temperature fell within 1◦ C of the actual temperature of a mussel. Reproduced from Helmuth (1998) with permission from the Ecological Society of America.

For example, Porter and colleagues computed geographic variation in operative temperatures of various ectotherms on multiple continents (e.g., Grant and Porter 1992; Kearney and Porter 2004). At the other extreme, Bartlett and Gates (1967) calculated the range of operative temperatures on the surface of a single tree. Mathematical models not only predict operative temperatures, they also predict how evaporative cooling and metabolic heating might cause the body temperature to deviate from the operative temperature (Casey 1992; Helmuth 1998; Tracy 1976). Although mathematical models have provided tremendous insight regarding the factors that inﬂuence operative temperature, they are impractical for mapping thermal environments at a sufﬁcient resolution to understand selective pressures on

behavior and physiology (O’Connor and Spotila 1992). To compute operative temperature from a mathematical model, one must know the solar radiation, ground reﬂectivity, air temperature, ground temperature, and wind speed (see Fig. 2.11). The overwhelming task of obtaining such data for a large number of locations precludes a ﬁne-scale mapping. In most cases, researchers have computed operative temperatures for a limited number of microclimates, such as full sun and full shade (Christian and Bedford 1995; Seebacher and Grigg 1997; Van Damme et al. 1989). As we shall see in later chapters, this coarse-scale mapping of the environment obscures much of the heterogeneity that drives thermal adaptation. For this reason, many researchers have adopted a different approach to study thermal heterogeneity.

THERMAL HETEROGENEITY

(a)

(b)

(c)

(d)

31

Figure 2.12 Physical models of operative temperatures are constructed to mimic the radiative, conductive, and convective properties of an organism. Examples include a copper electroform of a lizard (a), a mussel shell ﬁlled with silicone (b), a copper tube designed to mimic a snake (c) and a moist agar model shaped like a toad (d). Because the agar model incorporates evaporative water loss, this type of model records wet operative temperatures. (b) Courtesy of Brain Helmuth; (c) courtesy of Steven Beaupre; (d) courtesy of Carlos Navas.

2.3.2 Physical models Physical models provide a convenient alternative to mathematical models for mapping thermal heterogeneity on very small scales (Bakken and Gates 1975). Physical models mimic the radiative, convective, and conductive properties of living organisms in the absence of physiological function. As such, the temperature of the model at its steady state equals the operative temperature (Bakken 1976, 1992). In theory, a physical model should have no heat capacity and would reﬂect a fairly instantaneous measure of operative temperature. In practice, good physical models of operative temperatures respond rather quickly to changes in heat ﬂux relative to the responses of real organisms. To date, researchers have constructed physical models from a wide range

of materials, including metallic electroforms, rubber hoses, and dried exoskeletons (Fig. 2.12). For organisms that lack physiological control over evaporative cooling, models with a wet surface have been used to estimate steady-state body temperatures more accurately (Bartelt and Peterson 2005; Navas and Araujo 2000). When properly constructed, a physical model can predict the body temperature of an organism within 1◦ C of its actual value (reviewed by Dzialowski 2005). The most accurate models consist of hollow metallic electroforms painted to match the reﬂectivity of the organism (Fig. 2.12a). Metallic objects of similar dimension but slightly different shape yield larger errors (Walsberg and Wolf 1996). In the rare event that solar radiation and convection contribute

32

T H E R M A L A D A P TAT I O N

little to the heat ﬂux of an organism, simple physical models constructed from commercial temperature loggers might adequately estimate operative temperatures (Vitt and Sartorius 1999). Nevertheless, one should always determine the accuracy of a physical model under a range of conditions that reﬂect those of the environment (Bakken 1976; Dzialowski 2005). Because most types of physical models require little time and money to construct, researchers can use these models to record operative temperatures on a much ﬁner scale than they have done with mathematical models (Bakken 1989). For example, one can estimate the frequency of operative temperatures by placing models at randomly determined points (e.g., Grant and Dunham 1990) or at intervals along transects (e.g., Grant 1990). Alternatively, one can place models in speciﬁc microenvironments and use the area of each microenvironment to infer the frequency of operative temperatures (e.g., Grant and Dunham 1988). Collectively, these sampling designs have shown that operative temperatures depend on the type of substrate (Adolph 1990), the slope of the ground (Helmuth and Hofmann 2001; Sartorius et al. 2002), the amount of shade (Christian and Weavers 1996; Hertz 1992), and the posture and orientation of the organism (Bakken 1989; Christian and Bedford 1996; Grant and Dunham 1988). For a given location, posture, and orientation, operative temperatures rise throughout the morning and reach a maximum near the middle of the day (Bashey and Dunham 1997; Dorcas and Peterson 1998; Hertz 1992; Sartorius et al. 2002; Schäuble and Grigg 1998). Diel cycles of operative temperatures are more pronounced in terrestrial microclimates that receive solar radiation than they are in shaded or aquatic microclimates (Brown and Weatherhead 2000; Christian and Weavers 1996). Despite these insights, the majority of researchers have sampled environments at very low spatial and temporal resolutions. In a review of published studies, Dzialowski (2005) noted a typical sampling design has included fewer than ﬁve physical models deployed for no more than 5 days! Thus, while physical models offer practical advantages over mathematical models, very few researchers have capitalized on their potential to rigorously map operative thermal environments.

2.3.3 Statistical models A statistical model offers a powerful compromise between a highly mechanistic mathematical model and a purely phenomenological physical model. As seen in Section 2.1, we can use mathematics to describe an operative temperature in terms of heat ﬂuxes between an organism and its environment. Equation 2.10 describes a precise relationship between many independent variables and the operative temperature. Nevertheless, we would need to know the values of these variables at very high spatial and temporal resolutions to map operative temperatures with this mathematical model. Although we cannot easily determine all of these variables at many points in the environment, we can obtain reasonable proxies for these variables at a high resolution. Given these proxies, we could capture a great deal of information about operative thermal environments from a statistical model (Hertz 1992; Howe et al. 2007; Sears et al. 2004). Ultimately, the power of a statistical model depends on the degree to which the independent variables relate mechanistically to operative temperature (or correlate with variables that do). Commonly used proxies include air temperature, wind speed, slope, aspect, and elevation, but some models also include categorical variables such as the type of substrate or vegetation (Box 2.2). Statistical models enable researchers to create detailed maps of operative thermal environments. For example, Sears and colleagues (2004) constructed a model that related slope, aspect, elevation, wind speed, and air temperature to the operative temperature of a small lizard. The independent variables were obtained from micrometeorological equipment and a digital elevation map. For some locations, operative temperatures were estimated with physical models (see Section 2.3.2). These point samples were used to parameterize the statistical relationship between the independent variables and the operative temperature. Once the model was parameterized, operative temperatures were predicted for an entire landscape at a series of times (Fig. 2.14). These thermal maps have a resolution of less than a meter and a mean error of less than 2◦ C! This scope and resolution of thermal mapping greatly exceed those achieved with mathematical

THERMAL HETEROGENEITY

33

Box 2.2 Modeling operative temperatures with artiﬁcial neural networks

or physical models alone. As we shall see in the chapters that follow, the mapping of ﬁne-scale heterogeneity plays a critical role in understanding the selective pressures on organisms.

the model’s ability to predict the independent data even though its ability to predict the training data continues to improve. At this point, training should cease and the model should perform well when applied to new data. In general, artiﬁcial neural networks outperform linear models at predicting operative temperatures. For example, general linear models overpredicted low operative temperatures or underpredicted high operative temperatures (Howe et al. 2007; Sears et al. 2004). A potential drawback of an artiﬁcial neural network model is that mechanisms remain obscured by the “black box” of the model. Nevertheless, many ecologists have found these models ideal for predicting phenomena when underlying mechanisms exist but cannot be precisely formulated (Lek and Guégan 2000).

Continuous

Input layer

Discrete

Identifying the most accurate statistical model to predict operative temperatures can be challenging because these temperatures depend on the interaction of many environmental variables. Traditional statistical approaches, such as multiple regression analysis, often yield poor approximations of nonlinear relationships (Lek et al. 1996). To solve this problem, two groups of researchers have turned to artiﬁcial neural networks (Howe et al. 2007; Sears et al. 2004). An artiﬁcial neural network discovers the nonlinear relationship between a set of independent variables and a dependent variable. These models offer great ﬂexibility and power because (1) the independent variables can be either continuous or categorical and (2) the relationships between these variables and the operative temperature can take on many different forms (e.g., linear, sigmoidal, and step functions). Parameterizing an artiﬁcial neural network involves a process that mimics learning by a brain (reviewed by Basheer and Hajmeer 2000; Lek and Guégan 1999). The basic structure of the model consists of three interconnected layers: an input layer, a hidden layer, and an output layer (Fig. 2.13). The input layer consists of independent variables, such as air temperature, wind speed, slope, aspect, and elevation. The output layer consists of the operative temperature. In between, a hidden layer transforms the data from the input layer to the prediction at the output layer. This transformation involves a set of coefﬁcients and thresholds that serve as the parameters of the model. The process of learning begins by randomly choosing the values of these parameters. The model then proceeds to train itself by comparing the predicted and observed temperatures and adjusting the parameters to improve the ﬁt. After many iterations of training, an artiﬁcial neural network will converge on an excellent description of the data. The trick is to avoid overtraining the model such that it provides a perfect description of the training data but a poor description of novel data. To avoid overtraining, one must periodically confront the model with independent data during the training process. Initially, training improves the ability of the model to predict both the training data and the independent data. But, eventually, training will decrease

time Ta sun2 sun5 wind height substrate bare broken thin thick

Hidden layer

Output layer

Tex

Figure 2.13 The structure of an artiﬁcial neural network designed to predict microenvironmental temperatures (Tex ) from a combination of continuous and discrete variables. The continuous variables in this model include the time of day (time), air temperature at 1.3 m (Ta ), wind speed at 1.5 m (wind), solar radiation over previous 2–5 minutes (sun2 and sun5), and height above the ground (height). Discrete variables include the type of substrate and the degree of ground cover (bare, 0–29%; broken, 30–69%; thin, 70–100% with visible substrate; thick, 70–100% with no visible substrate). Adapted from Bryant and Shreeve (2002) with permission from Blackwell Publishing.

2.4 Conclusions Since its derivation, the concept of an operative temperature has guided efforts to characterize thermal

34

T H E R M A L A D A P TAT I O N

(a)

(b)

(c)

(d)

58 46 34 21 9

Figure 2.14 Operative temperatures available to small lizards in the Mescalaro sand dunes (New Mexico, USA) change dramatically throughout the morning on a summer’s day (M. W. Sears, G. S. Bakken, M. J. Angilletta, and L. A. Fitzgerald, unpublished observations). (a–d) Each map depicts an operative thermal environment over an area of 460 × 320 m at a resolution of 0.5 m. (a) 7 a.m.; (b) 9 a.m.; (c) 11 a.m.; (d) 1 p.m. See plate 1.

environments from the organismal perspective. Spatial and temporal variations in solar radiation and air temperature give rise to changes in operative temperatures. All else being equal, operative temperatures decrease with increasing latitude or altitude. However, substantial heterogeneity of operative temperature exists within local environments, particularly terrestrial environments at higher latitudes. This local heterogeneity results from the physical structure of the habitat, as well as diel, seasonal, and annual changes in meteorological conditions. Spatial and temporal variations in operative temperatures undoubtedly impose selective pressures on the thermal sensitivities, thermoregulatory strategies, and life histories of organisms. Both the

magnitude and predictability of thermal change should determine the evolutionary impact. Diel and seasonal changes can be extreme, but they come with certainty. Stochastic variation among days and years complicates adaptation to predictable cycles. The relative magnitudes of predictable and unpredictable changes vary with latitude. Moreover, the size, mobility, and lifespan of an organism determine its perception of thermal heterogeneity. For example, a fruit ﬂy experiences mostly diel change within its lifetime, but a Galapagos tortoise experiences diel, seasonal, and annual changes. As we shall see in the coming chapters, the course of thermal adaptation depends on the way each organism perceives the thermal heterogeneity of its environment.

CHAPTER 3

Thermal Sensitivity

3.1 Patterns of thermal sensitivity Anyone familiar with physics and chemistry should be unimpressed by the discovery that life depends on temperature. Temperature constrains the rates of chemical reactions, and biochemical reactions are no exception (Hochochka and Somero 2002). These constraints lead to thermal sensitivities of function at cellular, systemic, and organismal levels (Rome et al. 1992). The thermal sensitivity of biochemical reactions places severe limits on life (Pörtner 2002). Prokaryotic cells function within the narrow range of −5◦ C to 110◦ C (Jaenicke 1991, 1993). Eukaryotic cells are even further restricted, tolerating temperatures between −2◦ C and 60◦ C (Tansey and Brock 1972). Some organisms can tolerate temperatures outside these ranges, but do not function normally during exposure to such extremes. Yet, each living organism comprises a unique set of reactions. Thus, while the chemistry of life restricts the thermal tolerances of organisms, no two organisms have exactly the same thermal restrictions. By altering the structures of the reactants or even the nature of the reactions, species have evolved a rich diversity of thermal reaction norms. As an example, consider thermal reaction norms for ﬁtness, which I refer to as ﬁtness curves.6 Barlow (1962) compared the ﬁtness curves of two species of aphids raised in controlled environments (Fig. 3.1a). The ﬁtness of Myzus persicae was greatest at 20◦ C 6

I avoid the term tolerance curve (Levins 1968) because the literature contains conﬂicting uses of this term. Originally, Levins (1968) deﬁned a tolerance curve as the relationship between an environmental variable and ﬁtness. Since then, this term has also been used to describe the relationship between an environmental variable and survivorship (Gilchrist 1995). The word tolerance can also imply the critical thermal limits to performance (Pörtner 2002; van Berkum 1988).

and decreased sharply around this maximum. In contrast, the ﬁtness of Macrosiphum euphorbiae was maximized over a wider range of temperatures (20−25◦ C). Relatively speaking, we could call M. persicae a thermal specialist and M. euphorbiae a thermal generalist. Similarly, ﬁtness curves varied greatly among ﬁves strains of rotifers (Fig. 3.1b). Differences among ﬁtness curves have been observed for many kinds of organisms. Such observations demand explanation because natural selection should inﬂuence the thermal sensitivity of ﬁtness more strongly than it does any phenotype. Although differences among ﬁtness curves may motivate studies of thermal adaptation, the concept of a ﬁtness curve offers thermal biologists very little advantage for understanding most biological phenomena. First, reaction norms for ﬁtness cannot be measured accurately for many organisms. Fitness is a function of lifetime reproductive success, which obviously cannot be measured more than once per individual. In the examples above, asexual reproduction by aphids and rotifers enabled researchers to estimate the ﬁtness curves of clones. But clones cannot be obtained for the majority of organisms under study. Second, biologists cannot easily measure the ﬁtness of organisms in ecologically relevant environments. In most environments, temperature varies temporally within generations, and we might need to incorporate this variation in the concept of a reaction norm to accurately predict ﬁtness (Fig. 3.2). Moreover, nature contains a plethora of biotic and abiotic factors that one cannot replicate in a laboratory, and these variables can interact with temperature to determine components of ﬁtness (Fig. 3.3). Consequently, we cannot

35

36

T H E R M A L A D A P TAT I O N

Relative fitness (% of maximum)

(a)

(b) 100

100

80

80

60

60

40

40

20

20 0

0 0

5

10 15 20 25 Temperature (°C)

30

14

35

16

18 20 22 Temperature (°C)

24

26

Figure 3.1 Fitness curves vary within and among species. (a) Between two species of aphids, Macrosiphum euphorbiae (dashed line) tolerated higher temperatures than did Myzus persicae (solid line). Data are from Barlow (1962). (b) The thermal sensitivity of ﬁtness varied tremendously among strains of parthenogenetic rotifers. Data are from Ricci (1991). In both studies, ﬁtness was deﬁned as the Malthusian parameter, r (see Box 3.3).

0.5 P < 0.10

Fitness (r)

0.4 0.3

P < 0.05 0.2 0.1 Constant temperatures Fluctuating temperatures

0.0 24

26

28

30 32 34 36 38 40 Mean temperature (°C)

42

44

Figure 3.2 Thermal reaction norms for ﬁtness sometimes differ when measured under constant and ﬂuctuating temperatures. In this example, ﬁtnesses measured at two constant temperatures deviated from ﬁtnesses measured in ﬂuctuating environments with the same mean temperatures. Fitness was deﬁned as the Malthusian parameter, r (see Box 3.3). Data are from van Huis et al. (1994).

be certain that ﬁtness curves expressed in a laboratory resemble those expressed in nature. Finally, and perhaps most importantly, we generally do not ponder questions about the evolution of ﬁtness, per se. Rather, we wish to understand the evolution of phenotypes that presumably contribute to an organism’s ﬁtness. In this chapter, I focus on thermal reaction norms for organismal performance, which are generally referred to as thermal performance curves (Huey and

Stevenson 1979). I deﬁne performance as any measure of an organism’s capacity to function, usually expressed as a rate or probability. Common measures of performance include locomotion, assimilation, growth, development, fecundity, and survivorship7 (Table 3.1). Less common functions, such as sensory perception (Samietz et al. 2006; Werner 1973) and immunological activity (Hung et al. 1997; Mondal and Rai 2001), can also be considered performances. All of these functions share an important property: they respond rapidly (and usually reversibly) to changes in temperature. We can capture the general characteristics of this response with several parameters (Fig. 3.4): (1) the thermal optimum (Topt ); (2) the thermal breadth (or performance breadth); (3) the thermal limits, referred to as the critical thermal minimum (CTmin ) and the critical thermal maximum (CTmax ); and the maximal performance (Pmax ). To study variation in performance curves, we must ﬁrst be able to quantify them objectively. Several approaches have been used to quantify performance curves. First, a mathematical function can be chosen and its parameters can be estimated by 7

Other authors have distinguished between curves for survivorship and curves for other performances. For example, Gilchrist (1995) referred to curves for survivorship as tolerance curves. Later, van der Have (2002) referred to the same concept as a viability curve. For simplicity, I chose to treat survivorship as I would any other measure of performance.

THERMAL SENSITIVITY

Growth rate (calories day–1)

1000

Cmax

37

C = 2000

500

0 1000

C = 1000

C = 500

500

0

5

10

15

5 20 Temperature (°C)

10

15

20

Figure 3.3 The amount of food available to a ﬁsh determines its thermal sensitivity of growth rate. Plots are shown for ﬁsh consuming 500, 1000, 2000, and maximal (Cmax ) calories per day. Reproduced from Elliot (1982) with permission from Blackwell Publishing.

Table 3.1 Common measures of organismal performance. Examples of each kind of performance reﬂect relevant measures for various taxa. The sources contain detailed descriptions of the procedures for measuring these performances. Performance

Examples

Source

Locomotion

Voluntary speed Maximal burst speed Maximal jumping distance Maximal sustainable speed Maximal distance (endurance)

Gilchrist (1996); Heckrotte (1983) Bennett (1980); McConnell and Richards (1955) Navas et al. (1999) Scribner and Weatherhead (1995); Wilson and Franklin (2000a) Angilletta et al. (2002b); Full and Tullis (1990); Ojanguren and Bra˜na (2000) Samietz et al. (2006) Greenwald (1974); Van Damme et al. (1991) Albentosa et al. (1994); Dutton et al. (1975); Warren and Davis (1967) Howe and Marshall (2002); Kaufmann and Pough (1982) Kudo et al. (2000) Avery et al. (1993); Harwood (1979); Nicieza et al. (1994); Waldschmidt et al. (1986) Boddy (1983); Lebedeva and Gerasimova (1985); Lefﬂer (1972) Båmstedt et al. (1999); Brett et al. (1969) Li (2002); Stanwell-Smith and Peck (1998) Lamb and Gerber (1985) Berger et al. (2008); Howe (1962) Denoel et al. (2005); Wilson (2005) Carrière and Boivin (1997) Berrigan (2000); Lefﬂer (1972) Selong et al. (2001)

Feeding/assimilation

Growth Development Reproduction

Survival

Mechanosensory ability Probability of capturing prey Ingestion of food/energy Absorption of water/minerals Photosynthesis Digestion/passage of food Change in length Change in mass/energy 1/duration of developmental stage % per day Egg production Mating success Oviposition rate Probability of survival per unit time Time until 50% mortality

ﬁtting a statistical model to a set of data (Huey and Stevenson 1979). Choosing among potential models can be difﬁcult, but tools for model selection can help when no a priori reason exists to choose a particular model (Box 3.1). Alternatively, others have

adopted a simpler approach based on the concept of a minimum convex polygon (Bauwens et al. 1995; van Berkum 1986). Regardless of the approach, an objective description of the performance curve enables us to determine its key properties.

38

T H E R M A L A D A P TAT I O N

Performance

Pmax

Performance breadth CTmax

CTmin

Topt Temperature Figure 3.4 A hypothetical performance curve shows the stereotypical optimum at an intermediate temperature. The thermal optimum (Topt ), performance breadth, critical thermal limits (CTmin and CTmax ), and maximal performance (Pmax ) are labeled. Adapted from Huey and Stevenson (1979) by permission from Oxford University Press.

Ascertaining the thermal limits of performance can be difﬁcult, particularly when performance cannot easily be measured at extreme temperatures. Often, one must use an indirect means of bounding a performance curve. For example, curves for locomotor performance are generally bounded by some measure of immobility. For lizards, the inability of an individual to right itself has been used to estimate the critical thermal maximum. For fruitﬂies, the inability to cling to a surface has been used for the same purpose (Berrigan 2000; Huey et al. 1992). Lutterschmidt and Hutchison (1997a) argued that the temperature at which muscles begin to spasm provides the best measure of the critical thermal maximum. Whatever the measure, one should be aware of the assumptions associated with using indirect estimates of thermal limits to deﬁne performance curves (Lutterschmidt and Hutchison 1997 b).

Box 3.1 Selecting the best model of a thermal reaction norm

Maximal sprint speed (cm s–1)

To compare performance curves, we must be able to describe them objectively. This description usually involves ﬁtting a statistical model to empirical data. Several models have been favored by researchers, but how do we know the best model to describe any particular set of data? Instinctively, we might ﬁt several models to the data and choose the model that describes the greatest amount of variation (i.e., the model with the highest r 2 ). Although this approach seems reasonable at ﬁrst, further thought raises a

serious issue. Literally thousands of models exist to choose from, and some of these models are so complex they would over-ﬁt the data. In other words, a very complex model would not only describe the pattern of thermal sensitivity but also the error in the data (Burnham and Anderson 2002). Consequently, conclusions drawn from this model could be erroneous. What we desire is a model that describes the thermal sensitivity without ﬁtting noise in the data.

Gaussian Exponentially modified Gaussian

200

150

100

50

0 0

10

20 30 Temperature (°C)

40

50

Figure 3.5 The relationship between temperature and sprint speed as described by two statistical models: a Gaussian model and an exponentially modiﬁed Gaussian model. A comparison of values of Akaike’s information criterion indicated that the Gaussian model provides a better description of the data. Data are the mean sprint speeds of 12 lizards. Adapted from Angilletta (2006) with permission from Elsevier. © 2006 Elsevier Ltd. continues

THERMAL SENSITIVITY

39

Box 3.1 Continued Information theory provides a more sophisticated approach to describe performance curves (Angilletta 2006). One starts with a set of hypothetical models. These models can be chosen for their general form or derived from some mechanistic understanding (e.g., Schoolﬁeld et al. 1981; Sharpe and DeMichele 1977). Then, one ﬁts each model to the data, obtaining the estimated parameter values and the residual sum of squares. Finally, one calculates the Akaike information criterion (AIC) for each of the models (Burnham and Anderson, 2002): AIC = −2L + 2K +

2K (K + 1) , N −K −1

(3.1)

where L equals the maximized log-likelihood value of the model, K equals the number of parameters (including the error term), and N equals the sample size. The maximized log-likelihood value of a model can be computed easily from the model’s residual sum of squares (Burnham and Anderson, 2002). The AIC estimates the information lost when using a particular model to describe the data. Therefore, the best the model in the set is the one with the lowest value of AIC.

Any survey of performance curves within or among species will reveal some variation in the thermal optimum, breadth, and limits of performance. Consider the performances curves shown in Fig. 3.6. For any type of performance, the shape or position of the curve varies among closely related organisms. For example, the lizard Takydromus wolteri sprints fastest at 30◦ C, but its congener Takydromus sexlineatus sprints fastest at 34◦ C (Fig. 3.6a). Likewise, two distinct types of performance need not be equally sensitive to changes in temperature; compared to survivorship, sprint speed proceeds well over a very narrow range of temperatures (Fig. 3.6f). Is such variation in performance curves adaptive? If so, how does this variation inﬂuence the ability of organisms to survive and reproduce in their environments? Many species (or genera) span large portions of the earth. Obviously, species that extend from the tropics to the poles must be capable of tolerating

The AIC differs from a simple measure of ﬁt, such as the r 2 , because it also depends on the model’s complexity (i.e., the number of parameters). The second and third terms in eqn 3.1 eliminate the bias in ﬁt associated with more complex models. By correcting this bias, we avoid choosing a model that overﬁts the data. To illustrate this point, consider two models that I used to describe the thermal sensitivity of locomotor performance (Fig. 3.5). Although a simple Gaussian model (r 2 = 0.70, K = 4, AIC = 45.2) describes much less variation than does an exponentially modiﬁed Gaussian model (r 2 = 0.98, K = 6, AIC = 71.4), information theory indicates the simple model provides a better description of thermal sensitivity (for details, see Angilletta 2006). This result should satisfy theorists, who often use a Gaussian function to study the evolution of performance curves (see Section 3.3). At the same time, this result should alarm empiricists, who often record performance curves that appear asymmetrical. Because the AIC depends on the sample size, a larger sample might have supported the more complex model. This example underscores the need to collect sufﬁcient data to accurately describe the thermal sensitivity of performance.

thermal extremes. But how do these species accomplish this feat? Do individuals at the extremes of the range achieve lower ﬁtness than individuals at the center of the range? Or have subpopulations adapted to thermal heterogeneity within the geographic range? Two distinct strategies would enable a species to occur over a broad geographic range. Populations at the extremes of the range could consist of specialists that perform best at their respective environmental temperatures. Alternatively, the global population could consist of generalists capable of functioning over a wide range of temperatures. Which of these strategies, if any, should we expect to evolve? And can this adaptation explain the diversity of performance curves observed among species? To answer these questions, we must consider the processes that deﬁne the thermal sensitivity of performance and the evolutionary costs and beneﬁts of altering these processes.

T H E R M A L A D A P TAT I O N

(a)

(b)

100 80 60 40 Takydromus wolteri Takydromus sexlineatus Takydromus septentrionalis

20 0 15

Specific growth rate (d–1)

(c)

20

25

30

35

40 (d)

0.9 0.8 0.7 0.6 0.5

Spain Finland Russia Germany

0.4 0.3 16

18

20

22

24

26

28

(f)

90 60 30 0

Koliella anatarctica Chlorella sacchrophila Chlorella sorokiniana 0

10

20

30

40

0.05 0.04 0.03 0.02 Coenagrion hastulatem Coenagrion puella

0.01 12

14

16

18

20

22

24

26

28

1.00 0.98

80

Daily survivorship

Maximal daily fecundity

120

0.00 10

30

(e) 100

150

0.06 Developmental rate (d–1)

Sprint speed (cm s–1)

120

Photosynthesis (µmol O2 h–1 mg–2)

40

60 40 20 0

18

Woodland Agricultural 20

22

24 26 28 30 Temperature (°C)

32

34

0.96 0.94 0.92 0.90 0.88 0.86

36

0.84

Bactrocera cucurbitae Bactrocera dorsalis Bactrocera latifrons Ceratitis capitata 14 16 18 20 22 24 26 28 30 32 34 Temperature (°C)

Figure 3.6 Thermal sensitivities of performance vary within and among species. Examples shown here represent common measures of organismal performance: (a) sprinting speeds in three species of lizards (Chen et al. 2003; Ji et al. 1996; Zhang and Ji 2004); (b) gross photosynthesis in three species of algae (Vona et al. 2004); (c) growth rates in four populations of Daphnia magna (Mitchell and Lampert 2000); (d) developmental rates (1/duration of larval stage) in two species of damselﬂies (van Doorslaer and Stoks 2005b); (e) fecundity in two populations of butterﬂies, Pararge aegeria (Karlsson and Van Dyck 2005); and (f) survivorship in four species of fruitﬂies (Vargas et al. 1996). Where shown, error bars are ± 2 SE. (e) Adapted with permission from the Royal Society.

3.2 Proximate mechanisms and tradeoffs What proximate factors determine the shape of a performance curve? The precise answer to this question almost certainly depends on the performance

in question. The asymmetrical performance curves for locomotion and growth differ markedly from the more symmetrical performance curves for survivorship (van der Have 2002). The answer also depends on the way in which we measure performance.

THERMAL SENSITIVITY

3.2.1 Thermal effects on enzymes (and other proteins) Hochachka and Somero (2002) brilliantly summarized the thermal constraints on the functions of enzymes (see also Fields 2001; Jaenicke 1991; Somero 1995). For any chemical reaction, temperature determines the proportion of reactants that possess the free energy required for reaction; as the temperature increases, more reactants exceed the energy of activation. Enzymes lower the energy required for activation and thus speed the reaction at any given temperature. The function of the enzyme requires both an initial conformation that can bind the substrate(s) and a change in this conformation that creates an ideal environment for the reaction. Because of the diversity of weak interactions that maintain the structures of proteins, either too high or too low a temperature will destroy the necessary conformation to bind substrate. Even a moderate drop in temperature alters the rate of conformational change, leading to inactivation at low temperatures well above that required to denature the enzyme. Thus, extreme temperatures impair enzymatic function and consequently slow biochemical reactions, leading to performance curves for speciﬁc

100 % of maximal activity

Chronic exposure to a temperature usually elicits a different level of performance than acute exposure (Christian et al. 1986; Kingsolver and Woods 1997). Nonetheless, for a particular measure of performance recorded with a particular methodology, we can ask which cellular and systemic factors dictate the thermal sensitivity. Do the same factors determine the thermal optimum and the critical thermal limits? Is the critical thermal minimum necessarily linked to the critical thermal maximum, or can these parameters evolve independently? Likewise, does physiology dictate a genetic correlation between the performance breadth and the maximal performance? As we shall see, modeling the evolution of thermal sensitivity demands tentative answers to these questions (see also Pigliucci 1996). Therefore, before we explore some evolutionary models, let us consider some hypothetical cellular mechanisms for the thermal sensitivity of performance.

41

80 60 40 20 0 20

25

30 35 40 Temperature (°C)

45

50

Figure 3.7 Thermal sensitivities of the activity of adenosine triphosphatase (ATPase) resemble sensitivities of organismal performance. Data shown here are for four species of lizards: Dipsosaurus dorsalis (solid circles), Gerrhonotus multicarinatus (open circles), Sceloporus undulatus (triangles), and Uma notata (squares). From Licht (1967). Reprinted with permission from AAAS.

enzymes that resemble those for whole organisms (Fig. 3.7). The substitution of amino acids can alter the thermal properties of proteins (Fields 2001; Jaenicke 1991). Because the active site of an enzyme requires very speciﬁc amino acids to promote covalent reactions, this region generally remains conserved during evolution. Instead, natural selection favors mutations that alter the conformational stability of the enzyme (Hochochka and Somero 2002; Marx et al. 2007). A more ﬂexible structure helps an enzyme to change shape faster during catalysis. However, this instability of conformation comes at a price; a ﬂexible enzyme spends a greater fraction of its time in a conformation that prevents the binding of substrate (Fig. 3.8). Hence, a ﬂexible enzyme could have a lower afﬁnity for its substrate. In general, an enzyme with greater conformational stability functions better at high temperatures, and an enzyme with less conformational stability functions better at low temperatures (Fields 2001; Somero 1995). Two kinds of enzymatic variants can underlie the diversity of performance curves (Somero 1995). A mutation that alters the structure of an enzyme can produce an orthologous allozyme that enhances performance at some temperatures but reduces performance at others. Duplication of genes can lead to the evolution of paralogous isozymes, which

42

T H E R M A L A D A P TAT I O N

Distribution of conformational microstates Warm-adapted ortholog

Cold-adapted ortholog

(a)

Binding competent

Binding competent

More ordered

Conformation

High

Binding ability

Low

Rate

Less ordered Low High

(b)

Binding competent

Binding competent

Figure 3.8 Temperature inﬂuences the distribution of conformational microstates for orthologs of a protein from warm- and cold-adapted species. (a) At a given temperature, a cold-adapted ortholog possesses a broader distribution of conformational microstates because of its ﬂexible structure. For this ortholog, a larger fraction of molecules exists in conformations that bind ligands poorly (or not at all). However, these less ordered conformations reduce energy barriers for catalytic conformational changes and, therefore, speed enzymatic activity. (b) At their respective environmental temperatures, the intrinsic difference in stability between warm- and cold-adapted orthologs leads to similar distributions of conformational microstates for both orthologs. Adapted from Somero (2003) with permission from Elsevier. © 2003 Elsevier Ltd.

would enable an organism to operate over a wider range of temperatures. First, two identical copies of the gene would exist, and later one or both of these copies could evolve to alter the ancestral function of the enzyme. If one of these isozymes evolved greater or lesser ﬂexibility, the pair would complement each other in a variable environment. However, unless additional resources are used to synthesize isozymes, an organism will still be unable to increase its performance in one range of temperatures without decreasing its performance in another.

Based on correlative evidence, we might conclude that performance curves respond to evolutionary changes in the structures of enzymes (Mitten 1997). For example, orthologous allozymes for lactate dehydrogenase, LDH-Ba and LDH-Bb , enable thermal specialization of swimming endurance in the common killiﬁsh, Fundulus heteroclitus (reviewed by Powers and Schulte 1998). The catalytic efﬁciency of LDH-Bb exceeds that of LDH-Ba at temperatures below 25◦ C, while the opposite is true at temperatures above 25◦ C. The function of LDH-B inﬂuences the level of ATP in erythrocytes, which regulates the afﬁnity of hemoglobin for oxygen. Accordingly, homozygotes for LDHBb sustained faster swimming speeds at 10◦ C than did homozygotes for LDH-Ba , but these genotypes swam equally fast at 25◦ C. At high temperatures, I presume the swimming performance of homozygotes for LDH-Ba would exceed that of homozygotes for LDH-Bb , but this comparison has not been made. Consistent with this presumption, LDH-Ba occurs more frequently in populations that inhabit warm environments than it does in populations that inhabit cold environments. Importantly, neither genotype outperforms the other at all temperatures. Similarly, variation in the performance curves for growth rate among genotypes of Atlantic salmon (Salmo salar) reﬂects the synthesis of paralogous isozymes (Rungruangsak-Torrissen et al. 1998). A common form of trypsin, TRP-2∗ 100, functions at high temperatures (> 10◦ C), whereas the variant TRP-2∗ 92 functions at low temperatures (< 9◦ C). Salmon that synthesized both forms grew fairly well in the range of 4−20◦ C. A second variant, TRP-1∗ 91, functions over a wide range of intermediate temperatures, such that individuals synthesizing both TRP-2∗ 100 and TRP-1∗ 91 grew faster in the range of 6−17◦ C than did individuals synthesizing both TRP-2∗ 100 and TRP-2∗ 92. These two patterns of isozyme expression represent thermal specialization and thermal generalization, respectively. Because individuals within populations of salmon differ in the number and concentration of isozymes synthesized under controlled thermal conditions (Rungruangsak-Torrissen et al. 1998), we could hypothesize that performance curves evolved by selection for the synthesis of particular isozymes.

THERMAL SENSITIVITY

Gel phase

Region of phase separation

Low temperature

Fluid, or liquid-crystalline, phase

43

Inverted hexagonal phase High temperature

Physiological temperature Membrane fusion

Figure 3.9 The structure of a membrane can shift from a ﬂuid phase at tolerable temperatures (center) to either a gel phase at lower temperatures (left) or an inverted hexagonal phase at higher temperatures (upper right). The last phase ultimately results in fusion of the phospholipid bilayer (lower right). Reprinted, with permission, from the Annual Review of Physiology (Hazel 1995), Volume 57. © 1995 by Annual Reviews (www.annualreviews.org).

Although the evolution of isozymes appears less common than the evolution of allozymes (Fields 2001), both types of biochemical evolution could alter performance curves.

3.2.2 Membrane structure Temperature also determines the movement and conformation of cellular membranes (Fig. 3.9; Hazel and Williams 1990). On one hand, relatively low temperatures slow the movements of phospholipids, resulting in a gel-like membrane. On the other hand, relatively high temperatures speed these movements too much, disrupting the laminar structure. Passive regulation, active transport, and enzymatic function within membranes depend on a semiﬂuid structure. Therefore, membranes perform the same balancing act between ﬂexibility and stability performed by enzymes. Because of this balancing act, physiological functions involving membranes proceed best within a

moderate range of temperatures (reviewed by Hazel 1995). Although most organisms can maintain the full integrity of their cellular membranes over a breadth of 10−15◦ C, the optimal temperature depends on the composition of their phospholipids (Hazel 1995; Hazel and Williams 1990). Saturated fatty acids provide less intrinsic ﬂuidity than do unsaturated fatty acids. Consequently, membranes composed of primarily saturated fatty acids function better at high temperatures than membranes composed of primarily unsaturated fatty acids. An organism can produce membranes suited for cold or warm environments by using the appropriate mixture of fatty acids (Fig. 3.10). Still, adaptation to one thermal extreme necessarily leads to maladaptation to the other extreme. Do thermal effects on membranes limit organismal performance less than or more than thermal effects on enzymes? Although this question seems reasonable, we would have difﬁculty separating

44

T H E R M A L A D A P TAT I O N

0.30 0.28 0.26

Anisotropy

0.24

3.2.3 Oxygen limitation

Pigeon

DPH at 20°C TPA at 20°C DPH at Tb TPA at Tb

Rat

Perch

0.22

Cichlid

0.20 0.18 Trout

0.16 0.14 0.12 –10

Notothenia 0

10

20

30

40

50

Body temperature (°C) Figure 3.10 Membranes of cold-adapted species (e.g., an Antarctic ﬁsh, Notothenia neglecta) are far more ﬂuid (i.e., less ordered) than membranes of warm-adapted species (e.g., the rat or the feral pigeon). Species adapted to intermediate temperatures (perch, trout, and cichlid) possess intermediate levels of membrane ﬂuidity. Membrane ﬂuidity was assessed using ﬂuorescent probes: 1,6-diphenyl-1,3,5-hexatriene (DPH) and trans-parinaric acid (TPA). A high level of anisotropy reﬂects a low level of membrane ﬂuidity. Reprinted, with permission, from the Annual Review of Physiology (Hazel 1995), Volume 57. © 1995 by Annual Reviews (www.annualreviews.org).

the functions of membranes and enzymes in real cells. By disturbing the ﬂuidity of a membrane, extreme temperatures likely slow the function of proteins embedded within the phospholipid bilayer (Else and Wu 1999; Wu et al. 2001, 2004). Such disturbances can also affect the function of intracellular enzymes that rely on substrates transported through the membrane. Finally, the permeability of membranes partly determines the cellular milieu in which enzymes must function (Somero 2003). Very likely, the interaction between membranes and enzymes limits the performance of organisms (see, for example, Almansa et al. 2003). With this in mind, we shall now consider a model of thermal tolerance that involves both enzymes and membranes, particularly those associated with aerobic metabolism.

Recently, Pörtner and colleagues (2000) proposed a model of thermal tolerance based on oxygen limitation. This model posits that the thermal limits of performance are set by the temperatures at which aerobic respiration fails to meet energetic needs (Fig. 3.11). At low temperatures, organisms become incapacitated by the inability of mitochondria to generate the ATP required for activity. At high temperatures, ventilation and circulation fall below the level required to supply the mitochondria with sufﬁcient oxygen. This model draws support from the close match between the temperatures that cause death and the temperatures at which organisms shift from aerobic to anaerobic respiration (but see Klok et al. 2004; reviewed by Pörtner 2001). If this model correctly identiﬁes the cause of critical thermal limits, adaptation to thermal extremes will require changes in mitochondrial, circulatory, and respiratory capacities. A greater density or capacity of mitochondria could lower the critical thermal minimum of an organism (Guderley 2004; Pörtner et al. 2000). Still, such adaptations of metabolic systems would impose a tradeoff. The current view assumes that adaptation to cold environments causes maladaptation to warm environments. This hypothetical tradeoff results from an increase in proton leaking and futile cycling within mitochondria, which increase energy consumption at rest (Pörtner et al. 1998, 2000). The higher resting rate of energy consumption will cause a parallel shift in the critical thermal maximum. This model also predicts a potential tradeoff between the performance breadth and the maximal performance. Although an increase in mitochondrial density would raise metabolic demands at rest, organisms can eliminate this cost by reducing the leakiness of membranes (Pörtner et al. 1998) or the activity of enzymes (Pörtner et al. 2000). Either of these strategies would limit aerobic scope, potentially causing a decrease in the capacity for performance. Thus, the model of Pörtner and colleagues links the thermal limits, performance breadth, and maximal performance in ways that resemble the tradeoffs associated with enzymatic function (see Section 3.2.1).

THERMAL SENSITIVITY

45

(a)

% Oxygen limited 100 aerobic scope (steady state)

Blood oxygenation Anaerobic metabolism

0 (b) Mo2 ∆max

Residual ATP production capacity + Proton leakage

0 (c)

Q

(cardiac + ventilatory output)

∆max

Qmax

Qrest 0 (d) Rate of aerobic performance

0 Temperature

3.2.4 Conclusions from considering proximate mechanisms Of the biochemical and physiological processes discussed in this section, which one really limits performance at the organismal level? The answer likely depends on the temporal scale of thermal heterogeneity. Fields (2001) noted that the temperature required to denature a protein can exceed the critical thermal maximum of the organism. Based on this point and other evidence, Pörtner concluded that oxygen limitation offers the most plausible

Figure 3.11 A model of thermal tolerance based on limits to aerobic capacity. The mismatch between oxygen demand and supply sets the window of thermal tolerance of metazoans. (a) The onset of anaerobic metabolism sets the critical thermal maximum. (b) At both low and high temperatures, the mitochondria cannot supply sufﬁcient ATP. At low temperatures, the inactivation of mitochondrial enzymes and membranes prevents the production of ATP. At high temperatures, the rising fraction of mitochondrial proton leakage produces a mismatch between the supply and demand of oxygen. (c) Accordingly, the thermal sensitivity of ventilation and circulation, especially the difference between maximal and resting outputs, characterizes limits to oxygen supply and aerobic scope. (d) Thermal sensitivities of mitochondrial, ventilatory, and circulatory functions combine to generate an asymmetric performance curve, with a thermal optimum for performance near the temperature that maximizes aerobic scope. Adapted from Pörtner et al. (2006, Physiological and Biochemical Zoology, University of Chicago Press). © 2006 by The University of Chicago.

explanation for the thermal tolerances of organisms (Pörtner 2001, 2002). This conclusion may be valid for stable environments where organisms must grow and reproduce during persistent exposure to extreme temperatures. In heterogeneous environments, however, organisms could use anaerobic respiration to endure transient exposures to temperatures outside the window for aerobic scope. Other lines of evidence suggest the function of enzymes as a limiting factor for performance. The induction of heat-shock proteins under thermal stress implies that some loss of enzymatic function could

46

T H E R M A L A D A P TAT I O N

occur within the natural range of temperatures, which presumably exceeds the critical temperature for aerobic scope. van der Have (2002) found a striking match between the predicted thermal

limit of enzymatic activity and the observed thermal limit of embryonic survival (Box 3.2). Still, I believe that many factors contribute to the decline in organismal performance at extreme temperatures,

Box 3.2 Proximate basis of thermal sensitivity of development Thermal performance curves for survivorship during embryonic or larval stages possess a shape that differs from a typical performance curve (van der Have 2002). Instead of being asymmetrical, as in Fig. 3.4, these curves tend to be more symmetrical with very sharp transitions between the performance breadth and the thermal limits (see Fig. 3.6f). According to van der Have, this relationship stems from the reversible inactivation of proteins that drive cellular division. The gradual inactivation of enzymes with increasing or decreasing temperature can result in a thermal threshold for transcription (Fig. 3.12). Undoubtedly, cellular division and differentiation during development requires transcription. If proteins involved in transcription cannot maintain their functional conformation, the embryo or larvae would ultimately perish. This argument was formalized in a model describing the inﬂuence of a single rate-limiting enzyme (or group of enzymes) on development. The upper and lower temperatures at which transcriptional rates should drop

were calculated from the Sharpe–Schoolﬁeld model: r (T ) =

0.5

Expected thermal limit (°C)

Rate or probability

transcription rate probability in active state

HA R

1 1 298.2 − T

(3.2)

where r equals the rate of development, T equals temperature in K, R equals the universal gas constant, 298.2 equals the temperature at which no inactivation occurs, HA equals the enthalpy of activation, Pa equals the probability that the enzyme will be in the active state, and ρ represents a ﬁtted parameter (Schoolﬁeld et al. 1981; Sharpe and DeMichele 1977). Using developmental rates for 14 species, van der Have estimated the temperatures at which activation and inactivation were equally likely (i.e., Pa = 0.5). Based on his model of the cell cycle, these temperatures would shut down transcription within cells. A striking match exists between the theoretical limits for development and the empirical limits for survivorship (Fig. 3.13). 40

1.0

ρTPa e 298.2

Ceratitis Dacus Drosophila Dendroctonus Lipaphis Myzus

30

20

10

0.0 0

10

TL

20

30

TH Temperature (°C)

40

50

Figure 3.12 Temperature affects the probability that an inducer activates the transcription of an operon. The rate of transcription drops sharply as the temperature approaches the point at which the transducer is equally likely to be active or inactive. Adapted from van der Have (2002) with permission from Blackwell Publishing.

0 0

10 20 30 Observed thermal limit (°C)

40

Figure 3.13 An excellent match exists between the observed thermal limits to survivorship and the expected thermal limits of development. Data are for 14 species of insects from six genera. A reference line of y = x is shown. Adapted from van der Have (2002) with permission from Blackwell Publishing.

THERMAL SENSITIVITY

Low-temperature specialist

High-temperature specialist

Generalist Figure 3.14 Specialist–generalist tradeoffs constrain variation in performance curves. An improvement in performance at low temperature requires a decrement in performance at high temperature, and vice versa. Thus, a specialist for low or high temperature would achieve greater maximal performance than a generalist.

especially given the coadaptation that occurs among different elements of thermal physiology (a topic explored in Chapter 7). Overall, the thermal limits to performance probably depend on a complex interplay between biochemical, cellular, and systemic designs. Regardless of the primary cause of thermal sensitivity, we can draw at least one conclusion about the potential constraints on performance curves. Most of the cellular mechanisms described above involve a tradeoff between performance at high temperatures and performance at low temperatures. For example, adaptations of enzymes and membranes to low temperatures generally result in maladaptations to high temperatures. Adaptations that enhance the aerobic capacity at low temperatures create an energetic sink at high temperatures. These biochemical constraints should give rise to genetic correlations between parameters of a performance curve. First, a genotype that performs well in cold environments should perform poorly in warm environments. In the context of a performance curve, a shift in the critical thermal minimum should require a parallel shift in the critical thermal maximum. Second, a genotype that functions over a broad range of temperatures (e.g., through the expression of multiple isozymes) should be a jack of all temperatures and a master of none (Huey and Hertz 1984). In other words, an increase in the performance breadth should cause a decrease in the maximal performance. We can classify these predicted correlations as specialist–generalist

47

tradeoffs (see Chapter 1), because an evolutionary increment in performance at one temperature causes an evolutionary decrement in performance at another temperature (Fig. 3.14). Given the cellular mechanisms described above, specialist–generalist tradeoffs should impose ubiquitous constraints on evolution.

3.3 Optimal performance curves To understand the tremendous variation in performance curves, let us ﬁrst turn to optimality models. In particular, we shall focus on two distinct models designed to predict the relationship between a performance curve and ﬁtness. Each model predicts the thermal optimum and performance breadth that maximize ﬁtness in a given environment, subject to certain constraints. As we shall see, these models differ in fundamental ways, including the manner in which performance contributes to ﬁtness (Box 3.3). The ﬁrst model considers the thermal sensitivity of survivorship, or other performances that relate directly to survivorship. The second model considers the thermal sensitivity of fecundity, or other performances that relate directly to fecundity. Nevertheless, both models share many key features, including a strict constraint imposed by specialist–generalist tradeoffs. By examining these models in turn, we shall see how slightly different sets of assumptions lead to very different predictions regarding the optimal performance curve.

3.3.1 Optimal performance curves: survivorship and related performances Building on the seminal work of Levins (1968), Lynch and Gabriel (1987) modeled the evolution of performance curves for survivorship (i.e., the probability of survival during a unit of time).8 Their 8

Lynch and Gabriel envisioned their focus to be the evolution of ﬁtness curves, which they termed tolerance curves. I avoid their interpretation and their vocabulary for good reason. They considered how performance responds to thermal heterogeneity within the life of an individual. Yet ﬁtness is a property of an individual that should be characterized by a single value. Here, I reinterpret their model to understand the evolution of thermal sensitivities of survivorship because this interpretation seems equally valid and more intuitive.

48

T H E R M A L A D A P TAT I O N

Box 3.3 Performance and ﬁtness How will performance curves evolve in variable environments? To answer this question, we must understand how the performance of interest relates to ﬁtness. We might deﬁne ﬁtness as the number of offspring produced by an individual during its life. But when estimating the mean ﬁtness of a genotype (or phenotype), we cannot simply record the fecundity of individuals because many individuals will not live to reproduce at all. Therefore, we must also consider the probability that individuals will survive to reproduce. For practical purposes, we can deﬁne ﬁtness as the rate at which a genotype increases within a population. This measure of ﬁtness depends on two demographic variables: age-speciﬁc survivorship and age-speciﬁc fecundity (Stearns 1992). These variables contribute differently to the ﬁtness of a genotype in stable and growing populations (Kozłowski 1992). In stable populations, ﬁtness equals the net reproductive rate (R0 ): ∞ R0 =

l (x )m(x )dx ,

(3.3)

x =α

where l (x ) equals survivorship at age x , m(x ) equals fecundity at age x , and α is the age at maturity. In a growing population, one must estimate ﬁtness another way because offspring produced early in life contribute more to the increase of a genotype than do offspring produced late

model makes predictions for an asexual organism with discrete generations, but the main predictions also hold for a sexual organism (Gabriel 1988). This model can also help us to understand the thermal sensitivity of any other performance that directly affects survivorship (e.g., sprint speed). The model was deigned to determine the optimal performance curve in an environment that varies stochastically over space and time. For simplicity, the body temperature of the organism was assumed to equal its operative environmental temperature. The operative temperature was described by a Gaussian probability distribution. Hence, the organism experienced the mean temperature most

in life. For a population with a stable age distribution, ﬁtness equals the Malthusian parameter (r ), which we can estimate with Euler’s equation: ∞ 1=

e −rx l (x )m(x ) dx .

(3.4)

x =α

Note that eqn 3.3 deﬁnes a special case of eqn 3.4, in which the rate of population growth equals zero (i.e., r = 0). Because temperature affects r and R0 differently (Huey and Berrigan 2001), we need to know the appropriate measure of ﬁtness for each population under investigation. Because many populations remain stable over long periods (Sibly et al. 2007), some modelers feel that R0 represents a superior measure of ﬁtness (see Kozłowski et al. 2004). Regardless of the measure we choose, the ﬁtness function reveals an important point about the relationship between performance and ﬁtness. Performance can determine either the survivorship or the fecundity of a genotype at age x . Because l (x ) equals the product of survivorship from birth to age x , performances that affect survivorship contribute multiplicatively to the ﬁtness of a genotype. But performances that affect fecundity contribute additively to the ﬁtness of a genotype. As we shall see in this chapter, the distinction between multiplicative and additive contributions to ﬁtness plays a critical role in understanding the evolution of performance curves (see also Gilchrist and Kingsolver 2001).

often and experienced extremely high and low temperatures rarely. To understand how temporal and spatial sources of variation were combined, imagine a genotype develops, grows, and reproduces in a particular subset of its environment; let us call this subset the microenvironment. This microenvironment has a mean (φs ) and a variance (Vtw ) of temperature. Assuming microenvironments vary, the mean temperature of each microenvironment, φs , is drawn from a distribution for the macroenvironment, which has its own mean (φt ) and variance (Vφ s ) of temperature. Finally, the distribution of temperatures for the macroenvironment, φt , can change over time. Consequently, two kinds of temporal heterogeneity must be considered:

THERMAL SENSITIVITY

49

0.5

z1

z2

Survivorship

0.4 0.3 0.2 0.1 0.0 –3.0

–1.5

0.0 1.5 Environmental temperature

variance within generations (Vφ tb ) and variance among generations (Vφ tb ). As we shall see, these sources of variation do not equally affect the optimal performance curve. Lynch and Gabriel modeled the performance curve as a Gaussian function (Fig. 3.15): 2 P(z1 , z2 |φ) = (2πz2 )−0.5 e[−(z1 −φ) /2z2 ] ,

(3.5)

where P equals the performance (i.e., survivorship), z1 equals the thermal optimum, and z2 determines the performance breadth. Because this function describes a probability density, the integral (or area under the curve) remains the same for all values of z1 and z2 . In other words, this function conveniently imposes a specialist–generalist tradeoff because an increase in performance at one temperature necessarily reduces performance at another. The values of z1 and z2 of an individual were determined by genotypic and environmental factors: z1 = g1 + e1

(3.6)

z2 = g2 + e2 .

(3.7)

and

In these equations, g1 and g2 represent genetic determinants of the thermal optimum and performance breadth, respectively, whereas and e1 and e2 represent deviations from the genetic values caused by the developmental environment. This developmental noise affects the optimal performance curve very

3.0

Figure 3.15 Lynch and Gabriel used a Gaussian performance curve to model the evolution of thermal sensitivity. This curve was determined by two parameters: a thermal optimum (z1 ) and a √ performance breadth ( z2 ). Adapted from Gabriel and Lynch (1992) with permission from Blackwell Publishing.

little, but adaptation would proceed more slowly when either e1 or e2 differed from zero (Lynch and Gabriel 1987). The ﬁtness of a genotype, w( g1 , g2 ), depends on two things. First, we must consider its ﬁtness in any particular microenvironment (φs ), subject to the temporal variance of temperature (Vφ tw ). Then, we must consider the probability that a genotype will experience that particular microenvironment, given the spatial variance of temperature (Vφ s ). Because the product of survivorships at all time intervals deﬁnes the survivorship to reproduction, ﬁtness in a particular microenvironment must be calculated as the geometric mean of performance over the duration of a generation (τ ):  

τ

1/τ P(g1 , g2 |φs )i 

.

(3.8)

i=1

The probability of encountering a particular microenvironment was deﬁned as: 2 f (φs |φt ) = (2πVφs )−0.5 e[−(φs −φt ) /2Vφs ] ,

(3.9)

where f (φs |φt ) equals the probability of φs given φt . By combining Eqns. 3.8 and 3.9, Lynch and Gabriel deﬁned the expected ﬁtness of a genotype, w( g1 , g2 |φt ), given the spatial and temporal

50

T H E R M A L A D A P TAT I O N

(a)

(b)

(c)

Optimal breadth (g2)

200

Vfs =

Vftb = 0

100

150 100

Vftb = 10

Vftb = 100

100

10

10

0.1

0.1

0.1 10

50

100

0 0

20

40

60

80

100 0 20 40 60 80 100 0 20 Variance of temperature within generations (Vφtw)

40

60

80

100

Figure 3.16 Both spatial and temporal variations in environmental temperature affect the optimal performance breadth. (a–c) Each plot shows the effect of temporal variation within generations (Vφ tw ) for different levels of spatial variation (Vφ s = 0.1, 10, or 100); (b The three plots differ in the level of temporal variation among generations (Vφ tb = 0, 10, or 100). For these analyses, Lynch and Gabriel assumed the existence of developmental noise in the thermal optimum, but no developmental noise in the thermal breadth. Adapted from Lynch and Gabriel (1987, The American Naturalist, University of Chicago Press). © 1987 by The University of Chicago.

heterogeneity of temperature:  1/τ +∞ τ w( g1 , g2 |φt ) = f (φs |φt )  P(g1 , g2 |φs )i  dφs , −∞

i=1

(3.10) In biological terms, ﬁtness equals the probability of surviving until reproduction at age τ . Lynch and Gabriel implicitly assumed that survivorship has no inﬂuence on fecundity (i.e., all genotypes produce the same number of offspring). Therefore, the genotype with the greatest probability of surviving until τ could expect the greatest ﬁtness. Using this ﬁtness function, Lynch and Gabriel calculated the optimal values of g1 and g2 for a combination of spatial and temporal heterogeneity in a single generation (Fig. 3.16a). Obviously, the optimal performance curve should be centered on the mean temperature (i.e., g1 = φt ). But does spatial and temporal variation affect the optimal breadth of the performance curve? Intuitively, the optimal performance breadth (g2 ) increases with increasing thermal heterogeneity. Two other results seem less intuitive. First, developmental noise in the thermal optimum favors a reduction in the performance breadth. This result occurs because developmental noise creates phenotypes with different thermal optima, which reduces the need for a genotype to have an inherently wide performance breadth. Second, spatial and temporal heterogeneity interact

to determine the optimal genotype; at high levels of spatial variation, even small levels of temporal variation favor moderate performance breadths (Fig. 3.16a). Lynch and Gabriel went on to consider the longterm ﬁtness of genotypes with different performances curves. In doing so, they considered the effects of thermal heterogeneity within and among generations on the optimal performance curve. Variation among generations was modeled such that φt at each generation was drawn from a Gaussian distribution with a variance of Vtb . In this case, the ﬁtness of a genotype corresponds to the geometric mean of performance within and among generations (n): 1/n  n w( g1 , g2 ) = lim  w(g1 , g2 |φt ) (3.11) n→∞

i=1

By substituting eqn 3.10 into eqn 3.11, Lynch and Gabriel extended their calculation of the optimal performance curve from a single generation to inﬁnitely many generations. Importantly, extreme variation among generations always favors a greater performancebreadththanwouldbefavoredwhenthemean temperature remains the same among generations (Fig. 3.16b and c). Especially wide performance breadths are favored when the temperature within generations varies little over space or time. Despite these results, variation within generations has a

THERMAL SENSITIVITY

greater impact on the optimal performance breadth than does variation among generations. In summary, we can derive three testable predictions from Lynch and Gabriel’s model. First, the performance breadth for survivorship should be roughly proportional to the thermal heterogeneity experienced during an organism’s life. Second, the widest performance breadths for survivorship should evolve in spatially homogenous environments where temperature varies both within and among generations. Finally, both developmental noise and spatial variation prolong the time required for the optimal performance curve to evolve.

3.3.2 Optimal performance curves: fecundity and related performances Like Lynch and Gabriel, Gilchrist (1995) modeled the optimal performance curve of an asexual organism with discrete generations. But Gilchrist chose a ﬁtness function that differed fundamentally from the ﬁtness function described in the preceding section.

Low 50 (a)

In the model of survivorship, ﬁtness was calculated as the geometric mean of performance over the life of the organism. In Gilchrist’s model, ﬁtness was calculated as the sum of performance over the life of an organism. Arguably, performances such as growth and developmental rates contribute to ﬁtness in this additive fashion. If so, we can use Gilchrist’s modeltounderstandhowcertainperformancecurves evolve by natural selection, particularly when the organism reproduces semelparously or when adult survivorship does not vary with age (see Box 3.3). Gilchrist envisioned an organism whose body temperature varied dielly and seasonally according to its environment. Diel variation in temperature was stochastic, but seasonal variation was deterministic; speciﬁcally, seasonal variation followed a sinusoidal function. Because the organism completed many generations per year, seasonal variation occurred primarily among generations. Using nine combinations of diel and seasonal variations, Gilchrist determined the optimal performance curve for different levels of thermal heterogeneity within and among generations (Fig. 3.17).

Variation within generations Medium

High (c)

(e)

(f)

(h)

(i)

Low

(b)

25 0 50 (d)

Medium

25 0 50 (g)

Variation among generations

Mean temperature (°C)

51

High

25 0 0

60

120

180 0

60

120

180

0

60

120

180

Day of season

Figure 3.17 Gilchrist modeled the evolution of thermal sensitivity under nine combinations of daily and seasonal variations in temperature. These combinations created different levels of thermal heterogeneity within and among generations. Variation within generations depended on stochastic daily changes (error bars) and the sinusoidal seasonal changes. Variation among generations depended entirely on seasonal changes. Adapted from Gilchrist (1995, The American Naturalist, University of Chicago Press). © 1995 by The University of Chicago.

52

T H E R M A L A D A P TAT I O N

Instead of using a Gaussian function, Gilchrist modeled the performance curve using an asymmetric function (e.g., Fig. 3.4), which was characterized by a performance breadth (B) and critical thermal limits:

of the frequency distribution and the performance curve:

P(Tb , B, CTmax )

= 4 eρ(Tb −CTmin ) − e[(ρB)−1.2ρ(CTmax −Tb )]

The long-term ﬁtness of a genotype must be deﬁned as the geometric mean of ﬁtnesses for all generations (n):  1/n  CT n max    W= fg (Tb ) · P(Tb , B, CTmax )  

(3.12) where Tb equals the body temperature, B equals the range of temperatures bounded by the CTmin and the CTmax , and ρ determines the area under the curve. To impose specialist–generalist tradeoffs, Gilchrist maintained a constant area under the performance curve by setting the value of ρ according to the other variables. As mentioned above, the ﬁtness of a phenotype equaled the sum of performance throughout life. Because the temperature varied within and among days, the optimal strategy maximizes the sum of performance given the frequency distribution of temperatures in the current generation (fg ). Thus, the ﬁtness for a single generation equals the product

fg (Tb ) · P(Tb , B, CTmax ).

g=1 Tb =CTmin

(3.14) Using eqn 3.14, Gilchrist determined the optimal performance curve in the nine environments shown in Fig. 3.17. The majority of environmental conditions favor a thermal specialist with the smallest possible performance breadth (Fig. 3.18). The thermal optimum of this specialist should roughly equal the most frequent body temperature. This result makes perfect sense when the temperature remains relatively constant over time. But, surprisingly, an

High

(a) 10

(b) 1

(c) 0.2

(d) 2

(e) 1

(f) 0.2

(g) 1

(h) 0.5

(i) 0.2

Low

40

Variation among generations Medium High

30 Maximal temperature (°C)

(3.13)

Tb =CTmin

Variation within generations Medium

Low 50

CT max

w=

20 50 40 30 20 50 40 30 20 0

10

20

30 0

10

20

30 0

10

20

30

Performance breadth (°C) Figure 3.18 Most patterns of thermal heterogeneity explored by Gilchrist (1995) favor the evolution of thermal specialists. These ﬁtness landscapes correspond to patterns of thermal heterogeneity shown in Fig. 3.17. Contour lines delineate the ﬁtness of the various performance curves; the interval of the contour lines resides to the right of each plot’s label (e.g., a 10). The triangle on each ﬁtness landscape marks the optimal performance curve. Adapted from Gilchrist (1995, The American Naturalist, University of Chicago Press). © 1995 by The University of Chicago.

THERMAL SENSITIVITY

extreme specialist also achieves the greatest ﬁtness when temperature varies moderately or even greatly within generations. This conclusion holds as long the variation among generations remains low. In fact, the optimal performance breadth tightly follows the ratio of variation within generations to variation among generations. Specialists are more ﬁt than generalists when the variation within generations equals or exceeds the variation among generations (Fig. 3.18). The cause of this result should be obvious; a thermal specialist would become extinct during a generation when temperatures fall outside its performance breadth. Keep in mind, this model does not consider how spatial variation in temperature might preserve such genotypes. Gilchrist went on to consider how a bimodal distribution of temperatures would affect the optimal performance curve. Arguably, this extension of the model more accurately represents the thermal heterogeneity of most terrestrial environments and certain aquatic environments. The optimal performance curves were surprisingly similar to those of the initial model; extreme specialists possessed the greatest ﬁtness unless variation among generations exceeded variation within generations. Nevertheless, we should expect the thermal optimum for performance to be centered on the modal temperature instead of the mean temperature. Because of the asymmetry of the performance curve, a thermal optimum equal to the upper mode confers greater ﬁtness than a thermal optimum equal to the lower mode. If temperature does not vary, a thermal optimum equal to the lower mode confers the same ﬁtness as a thermal optimum equal to the upper mode. We shall return to these points in Chapter 7.

3.3.3 Contrasting the two models We have analyzed two models that predict optimal performance curves in variable environments. Armed with these models, we can try to understand the diversity of performance curves described in Section 3.1. Before doing so, let us brieﬂy compare and contrast the ideas captured by these models. Although both models consider a similar problem, each represents a unique set of assumptions about the nature of thermal heterogeneity, the constraints

53

on the phenotype, and the relationship between the phenotype and ﬁtness. These assumptions deﬁne the environmental, phenotypic, and ﬁtness functions of each model (Table 3.2). Not surprisingly, different assumptions lead to different predictions. In particular, the models provide radically different perspectives on the importance of temporal heterogeneity. Lynch and Gabriel’s model predicts that the performance breadth will be directly related to the thermal heterogeneity within generations. In contrast, Gilchrist’s model predicts that the performance breadth will be more affected by the relative amounts of thermal heterogeneity within and among generations. While these models differ in several ways, one seems most relevant to this result: the way in which performance contributes to ﬁtness. Therefore, we should expect to observe broader thermal breadths of survivorship in organisms subject to greater variation in temperature. At the same time, we should not expect to see this same relationship when comparing thermal breadths of other performances. Despite this discrepancy, both models predict that the optimal performance curve will enable maximal performance at the most frequent body temperature (or very close to this temperature). Therefore, support for either model would depend on the degree to which the thermal optima for various performances match the mean or modal temperatures. For example, if we compared genotypes from warm and cold environments, we should observe a pattern similar to the one shown in Fig. 3.19a. Alternatively, we might observe only partial adaptation to local environments (Fig. 3.19b), which would weakly support these models. Finally, we might observe a difference in performance curves between genotypes that provides no support for these models. In particular, both models would be refuted if genotypes from one environment performed as well as or better than genotypes from other environments at all temperatures. I refer to such genotypes as masters of all temperatures. When these masters of all temperatures come from the colder environment (Fig. 3.19c), the pattern of intraspeciﬁc variation is called countergradient variation because the genetic effects on performance counter the environmental effects on performance (Conover and Schultz 1995). When these masters

54

T H E R M A L A D A P TAT I O N

Table 3.2 A comparison of two models of optimal performance curves. Major assumptions underlying the environmental, phenotypic, and ﬁtness functions of each model are listed. Model component

Lynch and Gabriel’s model

Gilchrist’s model

Environmental function

• Temperature is normally distributed.

• Temperature is either normally distributed or bimodally distributed. • Mean temperature varies systematically among generations. • Mean temperature does not vary spatially. • Body temperature equals environmental temperature. • Temperature does not depend on other environmental variables. • The performance curve is skewed left. • The integral of the performance curve remains constant. • Developmental noise does not occur.

Phenotypic function

Fitness function

• Mean temperature varies stochastically among generations. • Mean temperature varies spatially. • Body temperature equals environmental temperature. • Temperature does not depend on other environmental variables. • The performance curve is symmetrical. • The integral of the performance curve remains constant. • Both genes and the environment (i.e., developmental noise) affect the performance curve. • The performance curve cannot acclimate to temperature. • Fitness for a single generation equals the product of performance over time. • Fitness over many generations equals the geometric mean ﬁtness.

of all temperatures come from the warmer environment (Fig. 3.19d), the pattern is called cogradient variation. Neither pattern should occur when the performance curves evolve according to specialist– generalist tradeoffs.

3.4 Using models to understand natural patterns Many species have large geographic ranges, spanning broad latitudinal or altitudinal gradients. As the range of a species expands, populations at the leading edge encounter novel thermal environments. In general, populations at high latitudes or altitudes experience colder environments than do populations at low latitudes or altitudes (Chapter 2). Populations at different latitudes or altitudes also experience different amounts of temporal variation in temperature. Based on the models described in the preceding section, we should expect very different selective pressures over the range of a species. Provided sufﬁcient genetic variation and limited gene ﬂow exist, performance curves would diverge

• The performance curve cannot acclimate to temperature. • Fitness for a single generation equals the sum of performance over time. • Fitness over many generations equals the geometric mean ﬁtness.

among populations (Grether 2005; Kawecki and Ebert 2004). Therefore, performance curves should vary predictably along latitudinal and altitudinal clines. To make useful predictions about geographic variation in performance curves, we must ﬁrst consider how the performance of interest contributes to ﬁtness. Consider the ﬁve performances most commonly measured: survivorship, locomotion, development, growth, and reproduction. How could each of these performances affect the ﬁtness of an organism? The effect of survivorship is obvious; survivorship to maturation and survivorship between successive reproductive events factor heavily in calculations of ﬁtness (see Box 3.3). Furthermore, Lynch and Gabriel’s model was speciﬁcally designed to predict the evolution of performance curves for survivorship (see Section 3.3.1). Using this model, we could make two predictions about the thermal sensitivity of survivorship: (1) colder environments, such as those found at high latitudes and altitudes, should lead to the evolution of lower thermal optima and (2) more variable environments, such as those found

THERMAL SENSITIVITY

(a)

Shift in thermal optimum (strong support for theory)

Performance Shift in thermal breadth (weak support for theory)

(d)

Performance

Performance

Temperature

Cogradient variation (no support for theory)

Hot source environment

Hot source environment Cold source environment

Hot source environment

Temperature

Temperature (b)

Countergradient variation (no support for theory)

Cold source environment

Hot source environment

Performance

Cold source environment

(c)

55

Cold source environment

Temperature

Figure 3.19 One of four patterns could emerge when comparing thermal performance curves between populations. (a) Evidence that thermal optima have diverged between populations strongly supports theory. (b) Evidence that performance curves have diverged and a specialist–generalist tradeoff occurred weakly supports theory. (c and d) Countergradient or cogradient variation does not support theory because thermal optima have not diverged and a specialist-generalist tradeoff has not occurred.

at high latitudes, should lead to the evolution of wider performance breadths. Now, consider the remaining performances. A greater capacity for locomotion can enhance foraging (Greenwald 1974), dominance (Robson and Miles 2000), and survivorship (Christian and Tracy 1981; Husak 2006; Jayne and Bennett 1990). If locomotor performance contributed signiﬁcantly to survivorship, we should expect its thermal sensitivity to also vary according to Lynch and Gabriel’s model. If locomotor performance contributes to ﬁtness more through fecundity than through survivorship, we should expect its thermal sensitivity to vary according to Gilchrist’s model. The same would be true for development, growth, and reproduction, which seem to contribute additively to ﬁtness. The sum

of development over time determines the developmental stage at the end of that time; reproduction generally requires developmental transitions among embryonic, larval, and adult stages. Likewise, the sum of growth over these stages determines the body size at maturation, which strongly inﬂuences the fecundity of an organism (Roff 2002). Finally, reproductive performance contributes directly to fecundity. When performance contributes to ﬁtness via its effect on fecundity rather than its effect on survivorship, we should expect slightly different patterns of variation along thermal clines. Although the thermal optimum should still decrease with increasing latitude or altitude, the performance breadth should vary only under certain conditions. Speciﬁcally,

56

T H E R M A L A D A P TAT I O N

geographic variation in the performance breadth should depend on the relative magnitude of intraand intergenerational variation in temperature (see Section 3.3.2). Obviously, these two components of variance depend on the lifespan and habitat of an organism. Short-lived aquatic species, such as Daphnia magna, experience relatively stable temperatures within generations because of the high heat capacity of water, but experience more variable temperatures among generations because of seasonal changes in water temperature. In such species, performance breadths for development, growth, and reproduction should increase with increasing latitude. For long-lived species (≥ 1 year), and particularly those in terrestrial habitats, performance breadths should remain relatively narrow over a latitudinal range. Armed with these speciﬁc predictions for real environments, we can see how well optimality models explain intra- and interspeciﬁc variation in performance curves. To keep things organized, we shall concentrate on the ﬁve performances described above, beginning with survivorship and ending with reproduction.

3.4.1 Survivorship The mere existence of organisms in extreme thermal environments, where most cannot survive, implies such organisms have adapted to these conditions (Clarke 1991). At one extreme, Antarctic ﬁsh swim in water cold enough to kill other ﬁsh (Wilson et al. 2001). At the other extreme, certain bacteria dwell in waters that would boil virtually any organism (Brock 1994). As assumed by current models, species adapted to extreme temperatures usually function poorly at moderate temperatures (see, for example, Wilson et al. 2001). Adaptation to thermal extremes seems to have come with a cost. This kind of qualitative support for the theory is reassuring, but the real utility of a theory comes from its ability to make predictions over a broad range of conditions. To determine whether current models provide this power, we must see how well they can predict intra- and interspeciﬁc variation in performance curves. Ideally, we would use direct comparisons of performance curves for survivorship to evaluate the

predictions of Lynch and Gabriel’s model. Surprisingly, such data are limited (Table 3.3). Although many researchers have documented relationships between temperature and survivorship (see review by Angilletta et al. 2004a), they have generally done so for one or two populations at a time. The paucity of well-designed comparative studies prohibits general inferences about the evolution of these performance curves; however, available studies paint a fuzzy picture. Moore (1940) made the ﬁrst effort to formally relate the thermal breadth of survivorship to the range of environmental temperatures. His summary of published data suggested aquatic animals possess much narrower breadths for survivorship than do terrestrial animals. More recently, at least one intraspeciﬁc analysis of survivorship supported the predicted relationship between thermal heterogeneity and performance breadth; in the beetle Hyperaspis notata, the performance breadth of a strain from high latitude was broader than that of a strain from low latitude (Dreyer et al. 1997). In the vast majority of intraspeciﬁc comparisons, however, little support can be gleaned for our predictions. Comparisons of survivorship over narrow (Drent 2002; Lankford and Targett 2001) or wide ranges (Ayres and Scriber 1994; Otterlei et al. 1999; Reisen 1995) of temperatures yielded no covariation between thermal heterogeneity and performance breadth. In fact, two latitudinal comparisons of tadpoles from the same species yielded different results. For a pair of populations of Rana temporaria (Stahlberg et al. 2001), tadpoles from a cold environment survived better at a low temperature than did tadpoles from a warm environment. For another pair of populations of the same species (Merila et al. 2000), tadpoles from the colder environment survived worse at a low temperature and better at a high temperature! These more recent comparisons within species and genera ﬂy in the face of Moore’s early study. Despite the inability of Lynch and Gabriel’s model to explain variation in performance breadth, the model still could explain variation in thermal limits of survivorship (reviewed by Garland and Carter 1994; Vernberg 1962). Many studies report direct measures of the thermal limits or indirect measures of thermal tolerance (Table 3.4). Indices of thermal tolerance include the temperature at which

THERMAL SENSITIVITY

57

Table 3.3 Comparative studies of the thermal sensitivity of survivorship. In most cases, individuals were raised from birth in a common garden (CG = Y), but in a few cases individuals either were acclimated to identical conditions for some number of days or were not acclimated (CG = N). Organism

Cline

CG

Measure(s)

Pattern

Source

No difference between populations

Ayres and Scriber (1994)

Canadian tiger swallowtail (Papilio canadensis) Atlantic croaker (Micropogonias undulatus) Common frog (Rana temporaria)

Latitude

Y

Survivorship at 12–30◦ C

Latitude

14 days

Survivorship at 1–5◦ C

No difference between populations

Lankford and Targett (2001)

Latitude

Y

Survivorship at 10–20◦ C for 15 days

Stahlberg et al. (2001

Common frog (Rana temporaria)

Latitude

Y

Survivorship of larvae at 14◦ C and 22◦ C

Beetle (Hyperaspis notata)

Latitude

Y

Survivorship at 15–34◦ C

Baltic clam (Macoma balthica) Mosquito (Culex tarsalis)

Latitude

Y

Survivorship at 10–20◦ C

Latitude

Y

Survivorship at 15–35◦ C

Latitude

Y

Survivorship at 4–14◦ C

Larvae from a cold environment survived better at 10◦ C than larvae from a warm environment. Larvae from a cold environment survived better at 22◦ C and worse at 14◦ C than larvae from a warm environment. B80 was broader for the strain from high latitude than it was for the strain from low latitude. No difference between populations Survivorship of females from a warm environment was higher at 20◦ C and 25◦ C than survivorship of females from a cold environment. No thermal or population trends

Atlantic cod (Gadus morhua)

Merila et al. (2000

Dreyer et al. (1997)

Drent (2002) Reisen (1995)

Otterlei et al. (1999)

Notes: B80 , 80% performance breadth.

animals lose motor function (knockdown temperature or critical thermal maximum), the duration of exposure to an extreme temperature before loss of function (knockdown time), and the time required to recover from exposure to an extreme temperature (recovery time). Because these measures are quick and easy to make on large numbers of individuals, many data exist to evaluate the predictions of optimality theory. We might deduce much about survivorship from these indirect measures of thermal tolerance. For example, we could interpret the knockdown temperature as an acute thermal limit to survivorship (Cooper

et al. 2008); in other words, survivorship equals zero at the knockdown temperature. Knockdown and recovery times probably correlate with the thermal limit to survivorship under chronic exposure, but should be interpreted with caution (Hoffmann et al. 2003). Certain geographic patterns of thermal tolerance make sense in light of Lynch and Gabriel’s model (Table 3.5). Early studies of reptiles and amphibians focused on variation in the critical thermal maximum, which arguably correlates with the acute thermal limit of survivorship. In some taxa, the CTmax tends to be greater for individuals that inhabit

58

T H E R M A L A D A P TAT I O N

Table 3.4 Common indices of the tolerance of extreme temperatures. Examples were chosen to illustrate applications in diverse taxa. Indices of freeze tolerance (e.g., supercooling point) were deliberately omitted because they do not reﬂect adaptation to temperature per se (Clarke 1991). Measure

Description

Example

Lethal temperature

The temperature that causes the death of 50% of the individuals in a sample (abbreviated as LT50 ) The time required for 50% of the individuals in a sample to die from exposure to a temperature (also abbreviated as LT50 , but should not be confused with the lethal temperature) Minimal or maximal temperature at which motor function ceases or muscular spasms occur (often used to place an upper or lower bound on a performance curve) Probability of survival following a brief exposure to a high temperature Probability of survival following a brief exposure to a low temperature The minimal temperature that causes an individual to lose motor function during heating The time required for an individual to lose motor function during exposure to a high temperature The maximal temperature that causes an individual to lose motor function during cooling The time required for an individual to recover from loss of motor function following exposure to a low temperature

Stillman and Somero (2000)

Lethal time

CTmin or CTmax

Heat-shock survivorship Cold-shock survivorship Knockdown temperature Knockdown time Chill-coma temperature Chill-coma recovery

warmer environments (Hertz 1979; Hertz et al. 1979; Licht et al. 1966a; Miller and Packard 1977; but see Spellerberg 1972). In other taxa, both the CTmax and CTmin decrease with increasing latitude (Snyder and Weathers 1975). In both situations, the thermal limits of performance appear to have shifted in the direction of the environmental temperature. Still, other patterns make no sense. For example, several investigators found no divergence in the CTmax or lethal limits along latitudinal clines (Gaston and Chown 1999; Klok and Chown 2005; Winne and Keck 2005). Overall, lower thermal limits seem to vary more closely with latitude than do upper thermal limits, which could stem from the steeper latitudinal cline in minimal environmental temperature (Ghalambor et al. 2006). The CTmin of dung beetles decreased with increasing latitude, despite no cline in CTmax (Gaston and Chown 1999). Furthermore, Castenada ˜ and colleagues (2005) observed a very strong relationship between the minimal environmental temperature and the chill-coma recovery time among four populations of woodlice. Importantly, most of these studies involved a small number of populations or species and included no phylogenetic control for pseudoreplication (see Box 1.5).

Berrigan (2000)

Gaston and Spicer (1998); Miller and Packard (1977); Winne and Keck (2005) Sorensen et al. (2005) Gilchrist et al. (1997) Huey et al. (1992) Huey et al. (1992) Gibert and Huey (2001) Castaneda et al. (2005)

More rigorous studies included larger samples or phylogenetic control. Stillman and Somero (2000) focused on latitudinal variation in the upper lethal temperature of crabs in the genus Petrolisthes. They found that crabs from warm environments survived acute exposures to higher temperatures than did crabs from cold environments (Fig. 3.20a). The same result was obtained when independent contrasts were used (Fig. 3.20b), indicating that adaptation of the thermal limit of survivorship has occurred within this genus. Despite this strong support for the current theory, the most comprehensive intraspeciﬁc study to date provided little support. AddoBediako and colleagues (2000) compared lower lethal temperatures and critical thermal maxima among species of insects. They found no evidence that the CTmax varied latitudinally. Although lower lethal temperatures decreased at the northern latitudinal extreme, this decrease in the mean resulted from an increase in the variance. Many species from very high latitudes exhibited lower lethal temperatures as high as those of species from low latitudes. A more comprehensive picture of the evolution of thermal tolerance has emerged from comparisons among populations and species of Drosophila

THERMAL SENSITIVITY

59

Table 3.5 Comparative studies of thermal tolerance, including lethal thermal limits (LT50 , CTmin , or CTmax ), heat resistance (knockdown temperature or time), and cold resistance (chill-coma temperature or recovery time). In most cases, individuals were raised from birth in a common garden (CG = Y), but in a few cases individuals either were acclimated to identical conditions for some number of days or were not acclimated (CG = N). Organism(s)

Cline

CG

Measure(s)

Pattern

Source

Algae Valonia utricularis

Latitude/longitude

Y

lower and upper lethal temperatures (LT50 s)

Lower and upper LT50 s were positively related to minimal and maximal environmental temperatures, respectively.

Eggert et al. (2003a)

Insects Drosophila buzzatii

Altitude

Y

Y

Altitude

Y

Knockdown time at 40◦ C

Latitude

Y

Loss of male fertility (CTmax )

Latitude

Y

Times to recover from exposures between –7◦ C and 2◦ C

Latitude

Y

Survival at –2◦ C for 2 hours

Latitude

Y

Recovery from 8 hours at 0◦ C

Latitude

Y

Knockdown time at 39◦ C

Latitude

Y

Latitude

Y

Time to recover from exposure to 0◦ C for 16 hours Loss of righting response (CTmin )

Drosophila serrata Drosophila spp.

Latitude

Y

Latitude

Y

Drosophila spp.

Latitude

Y

Knockdown time was negatively related to altitude; survivorships after heat and cold shocks were unrelated to altitude. Knockdown time was negatively related to latitude; in females, survivorship after heat shock was positively related to latitude (no relationship existed for males). Knockdown time was greater for populations from lower elevation. Tropical species had a higher CTmax than did temperate species. Temperate populations recovered faster than subtropical populations at low temperatures. Survival was positively related to latitude. Recovery time was negatively related to latitude. Knockdown time was negatively related to latitude. Recovery time was negatively related to latitude. CTmin was negatively related to latitude. Recovery time was negatively related to latitude. Cold tolerance was greater for species from higher latitudes, but heat tolerance was unrelated to latitude. Heat tolerances were greater for populations from lower latitudes.

Sorensen et al. (2005)

Latitude

Knockdown time at 37◦ C; survivorship after exposure to 40.5◦ C for 1 hour; recovery from 12 hours at 0◦ C Knockdown time at 40◦ C; survivorship after exposure to 40.5◦ C for 1 hour

Drosophila spp.

Latitude

Y

Drosophila melanogaster

Time to recover from exposure to 0◦ C for 8 hours Lower and upper lethal temperatures

Lethal time (LT50 ) at high temperature, resistance to knockdown, and knockdown temperature Time to recover from exposure to 0◦ C for 16 hours

Temperate species recovered faster than tropical species.

Sarup et al. (2006)

Sorensen et al. (2001) Rohmer et al. (2004) David et al. (2003)

Hoffmann et al. (2002) Hoffmann et al. (2002) Hoffmann et al. (2002) Ayrinhac et al. (2004) Gibert and Huey (2001) Hallas et al. (2002) Ohtsu et al. (1998)

Berrigan (2000)

Gibert et al. (2001) Continued

60

T H E R M A L A D A P TAT I O N

Table 3.5 Continued Organism(s)

Cline

CG

Measure(s)

Pattern

Source

Drosophila and Scaptodrosophila spp.

Latitude

Y

Lower and upper lethal temperatures

Kimura (2004)

Embryonopsis halticella Lycaena tityrus

Latitude

N

Lethal limits

Cold tolerance was greater for species from higher latitudes, but heat tolerance was unrelated to latitude; within species, cold tolerance was greater for strains from higher latitudes. No differences between populations

Altitude

Y

Time to recover from exposure to –20◦ C for 6 min; knockdown time at 47◦ C

Insecta

Latitude

N

CTmax and lower and upper lethal temperatures

Scarabaeidae

Altitude

4 days

Loss of righting response (CTmin ) and onset of spasms (CTmax )

Crustaceans Orchestia gammarellus

Latitude

51 days

Lethal temperature and lethal time

Petrolisthes spp.

Latitude

N

Upper lethal temperature (LT50 )

Porcellio laevis

Latitude/altitude

14 days

Times to recover from exposures between 0◦ C and 7◦ C

Altitude

4 days

Latitude/altitude

Cold tolerance increased with increasing altitude; heat tolerance was greater in populations from low and middle altitudes than in a population from high altitude. Lower lethal temperature was very low at high northern latitudes; upper lethal temperature and CTmax were unrelated to latitude. CTmin was negatively related to latitude; CTmax was unrelated to latitude.

Klok and Chown (2005) Karl et al. (2008)

Addo-Bediako et al. (2000)

Gaston and Chown (1999)

Individuals from a warm environment tolerated heat better than individuals from a cold environment. LT50 was positively related to maximal environmental temperature. Recovery time was positively related to minimal environmental temperature.

Gaston and Spicer (1998)

Loss of motor function (CTmax )

CTmax was negatively related to altitude.

Miller and Packard (1977)

N

Duration of survival at 46◦ C

Licht et al. (1966a)

Altitude

N

Anolis semilineatus Anolis gundlachi

Altitude

N

Altitude

N

Nerodia rhombifer Zootoca vivipara

Latitude

Y

Altitude

N

Onset of spasms and loss of righting response (CTmax ) Onset of spasms and loss of righting response (CTmax ) Onset of spasms and loss of righting response (CTmax ) Onset of spasms and loss of righting response (CTmax ) Loss of righting response (CTmin and CTmax )

CTmax was positively related to maximal ﬁeld body temperature. CTmax was negatively related to altitude. CTmax was negatively related to altitude. CTmax was negatively related to altitude. No relationship with latitude

Amphibians Pseudacris triseriata Reptiles Amphibolurus spp. Anolis krugi

No differences between populations

Stillman and Somero (2000) Castañeda et al. (2005)

Hertz (1979) Hertz (1979) Hertz et al. (1979) Winne and Keck (2005) Gvoždík and Castilla (2001)

THERMAL SENSITIVITY

(a) 42 40

8

38

6

(b) 1.4 2

1.2 1.0 Contrast for LT50

9 12 10 11

36 LT50(°C)

5 7

1 3 4

13

34

14 16

32

15

17

30 19

28

15

0.6 0.4

0.0

r 2 = 0.88

26 10

0.8

0.2

18

20

21

61

20 25 30 35 Maximal temperature (°C)

40

45

r 2 = 0.72

–0.2 0.0

0.5 1.0 1.5 2.0 2.5 Contrast for maximal temperature

3.0

Figure 3.20 Among species of porcelain crabs, the lethal temperature (LT50 ) was strongly related to the maximal temperature of the microenvironment. This relationship held for analyses of raw data (a) and independent contrasts (b). In the plot of raw data, each point represents the LT50 of a species: 1, Petrolisthes gracilis; 2, P. tridentatus; 3, P. armatus, northern Gulf of California; 4, P. armatus, Panama; 5, P. holotrichus; 6, P. hirtipes; 7, P. platymerus; 8, P. crenulatus; 9, P. sanfelipensis; 10, P. agassizii; 11, P. haigae; 12, P. galathinus; 13, P. granulosus; 14, P. cabrilloi; 15, P. cinctipes; 16, P. laevigatus; 17, P. violaceus; 18, P. tuberculatus; 19, P. manimaculis; 20, P. eriomerus; 21, P. tuberculosus. Adapted from Stillman and Somero (2000, Physiological and Biochemical Zoology, University of Chicago Press). © 2000 by The University of Chicago.

(reviewed by Hoffmann et al. 2003). Most studies to date accord with our theoretical predictions. For three species, intraspeciﬁc comparisons of chill-coma recovery revealed that tolerance of low temperatures has adapted to local climates. Flies of both D. melanogaster and D. serrata from high latitudes recovered faster than conspeciﬁc ﬂies from low latitudes (Ayrinhac et al. 2004; Hallas et al. 2002). Drosophila buzzatii exhibited a similar cline in chillcoma recovery over its altitudinal range (Sorensen et al. 2005). Variation among species of Drosophila can also be explained by the theory. Intuitively, tropical species of Drosophila require far more time to recover from chill coma than do temperate species (Fig. 3.21). A phylogenetic comparative analysis by Gibert and Huey (2001) revealed correlated changes in latitude and the lower thermal limit during the evolution of a major clade; as expected, species that encountered higher latitudes have evolved a lower chill-coma temperature. Collectively, these data provide strong evidence that thermal limits in species of Drosophila have adapted to local thermal environments. Nevertheless, the same data also indicate that adaptation to one thermal extreme

does not necessarily cause maladaptation to the other. For example, tolerances to heat and cold seem to have evolved in parallel within D. melanogaster (Hoffmann et al. 2002), but have evolved independently within D. buzzatii (Sorensen et al. 2005). Unfortunately, we do not know whether the apparent increases in the breadth of thermal tolerance have imposed any costs of maximal performance. In summary, measures of survivorship under chronic exposures to constant temperatures do not support the current theory, whereas measures of thermal limits during acute exposures partially support the theory. Generally, thermal optima do not decrease and performance breadths do not increase with increasing latitude as predicted by Lynch and Gabriel’s model. In contrast, measures of lethal temperatures, knockdown times, and recovery times often covary predictably with latitude and altitude. In some instances, latitudinal variation in the lower thermal limit occurs in the absence of latitudinal variation in the upper thermal limit, suggesting greater thermal heterogeneity at high latitudes has led to the evolution of thermal generalists. Perhaps measures of acute thermal tolerance relate more to

Number of observations

62

T H E R M A L A D A P TAT I O N

15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

Tropical species Temperate species

0

10

20

30

40 50 60 70 80 Recovery time (min)

90

100 110 120

phenotypes under selection in natural environments than do measures of chronic thermal tolerance.

3.4.2 Locomotion The thermal sensitivities of jumping, swimming, walking, and sprinting have been estimated for many species (Table 3.6). In amphibians, the thermal sensitivities of jumping and swimming vary predictably within and among species. Wilson (2001) compared jumping performances among populations of the marsh frog, Limnodynastes peronii, along both latitudinal and altitudinal clines in eastern Australia. As expected, frogs from cool environments jumped better at low temperature and worse at higher temperatures than did frogs from warm environments; the performance breadth also appeared to be broader in cooler environments, but measurements were not taken over a broad enough range to compare performance breadths quantitatively. A similar study was conducted by John-Alder and colleagues, who compared jumping performances among species of North American frogs; they observed broader performance breadths in species that bred in colder environments. Extensive comparisons of swimming performance have been made among populations and species of frogs in the Andes of South America (reviewed by Navas 2006). Navas (1996a) showed that four species of frogs from high elevations had lower thermal optima and broader performance breadths than did their congeners from low elevations. In two of the four

Figure 3.21 Tropical species took far longer than temperate species to recover from an exposure to 0◦ C for 16 hours. Data are means for 26 temperate and 48 tropical species of Drosophila. Adapted from Gibert et al. (2001) with permission from Blackwell Publishing.

cases, this expansion of the performance breadth seems to have come at a cost of maximal performance (Fig. 3.22), indicating a specialist–generalist tradeoff. Similar adaptations along altitudinal clines have been observed within and among other species of anurans (Navas 2006). For a combination of historical and logistical reasons, the most comprehensive comparisons of thermal sensitivities have focused on the maximal sprint speeds of lizards (reviewed by Angilletta et al. 2002a). Brieﬂy, one stimulates an individual to run along a linear track and records the greatest velocity achieved during the run. Usually, one records the fastest of several such runs as the maximal sprint speed. This procedure is repeated at different temperatures to construct a performance curve for each individual. An early study by Hertz and colleagues suggested the thermal sensitivity of sprint speed in an agamid lizard (Stellio stellio) had not adapted to thermal variation along an altitudinal cline. Later, more extensive investigations produced evidence of thermal adaptation. Among ﬁve species of Anolis, the thermal optimum for sprinting was correlated with the median ﬁeld body temperature, and the 95% performance breadth9

9

Although mathematical deﬁnitions of the performance breadth vary among models, empiricists typically estimate this parameter as the range of temperatures that enables performance that equals or exceeds some percentage of the maximal performance. The two most common estimates are the 95% performance breadth and the 80% performance breadth.

THERMAL SENSITIVITY

63

Table 3.6 Comparative studies of thermal performance curves. These exemplary studies were chosen for their reasonable samples along thermal clines (at least three populations per cline or multiple clines) and sufﬁcient resolutions of thermal performance curves (at least ﬁve temperatures). In most cases, individuals were raised from birth in a common garden (CG = Y), but in some cases individuals either were acclimated to identical conditions for some number of days or were not acclimated (CG = N). Performance

Organism(s)

Cline

CG

Pattern

Source

locomotion Sprinting

Anoles

Latitude/altitude

N

van Berkum (1986)

Sprinting Swimming

Starred agama Various genera of frogs

Altitude Altitude

N N

Jumping

Tree frogs

Latitude

N

Jumping

Striped marsh frog

Latitude/altitude

N

Rolling

Isopods

Latitude

N

Sprinting

Lizards

Latitude/altitude

N

Topt increased with increasing ﬁeld body temperature; B80 increased with increasing variation in body temperature. None Topt was lower for each congener at a higher altitude. B80 increased with increasing variation in body temperature. Topt increased with increasing environmental temperature. Topt increased with increasing environmental temperature. Topt and CTmax increased with increasing ﬁeld body temperature; CTmin was unrelated to Topt and CTmax .

Development Embryonic Common frog Larval Wood frog Larval Green frog

Latitude Altitude Latitude/altitude

Y Y Y

None Countergradient variation Countergradient variation

Larval

Latitude

Y

None

Assimilation/growth Wet mass Marine algae

Latitude/longitude

Y

Cell number

Bacteria

Latitude

Y

Dry mass

Trees

Latitude

Y

B80 increased with increasing range of environmental temperatures. Topt was related to water temperature. Topt was related to latitude.

Photosynthesis Photosynthesis

Pondweed Aquatic macrophytes

Latitude Latitude

Y N

none Topt was related to latitude.

Cell number Cell number

Symbiotic algae Marine algae

Latitude Latitude

Y Y

Topt was related to latitude. Topt did not vary, but B80 and thermal limits were widest for algae from the most stable environment.

Pea aphid

Hertz et al. (1983) Navas (1996a) John-Alder et al. (1988) Wilson (2001) Casta˜neda et al. (2004) Huey and Kingsolver (1993)

Laugen et al. (2003b) Berven (1982) Berven and Gill (1983; Berven et al. (1979) Lamb et al. (1987) Eggert et al. (2003a)

Hahn and Pockl (2005) Cunningham and Read (2003a) Pilon and Santamaría (2002) Santamaría and van Vierssen (1997) Sun and Friedmann (2005) De Boer et al. (2005)

Continued

64

T H E R M A L A D A P TAT I O N

Table 3.6 Continued Performance

Organism(s)

Cline

CG

Pattern

Source

Dry mass Energy

Water ﬂea Atlantic halibut

Latitude/longitude Latitude

Y 40–50 days

Mitchell and Lampert (2000) Imsland et al. (2000)

Length

Medaka

Latitude

Y

Length/mass

Silverside

Latitude

Y

Wet mass Reproduction number of infective juveniles

Arctic charr

Latitude

Y

Cogradient variation Countergradient variation Countergradient variation Countergradient variation none

Yamahira and Conover (2002) Larsson et al. (2005)

parasitic nematodes

worldwide

Y

none

Grewal et al. (1994)

Yamahira et al. (2007)

Note: Topt , thermal optimum for performance; B80 , 80% performance breadth; CTmin , critical thermal minimum; CTmax , critical thermal maximum.

Atelopus

30

*

*

20

A. sp. nov. A. muisca A. varius

10 Strokes in 10 s

*

*

*

20

Colostethus

30

10

C. subpunctatus C. flotator C. talamancae

0

0 0 30

10

20

30 30

Eleutherodactylus

*

20

0

10

20

Hyla

*

* 20

* 10

*

10 E. bogotensis E. diastema

0 0

10

30

20

30

H. labialis H. microcephala H. ebraccata

0 0

10

20

30

Body temperature (°C) Figure 3.22 The thermal sensitivity of swimming performance differed between congeners in four genera distributed along an altitudinal gradient. Solid and open symbols denote species from high and low altitudes, respectively. Adapted from Navas (1996a, Physiological and Biochemical Zoology, University of Chicago Press). © 1996 by The University of Chicago.

was correlated with the range of ﬁeld body temperatures (van Berkum 1986). An even larger study comprising 18 species of lizards representing four families lent additional support to the current theory

(van Berkum 1988). Among these species, the 95% performance breadth and the range of temperatures between the thermal limits (CTmax − CTmin ) were correlated with the range of ﬁeld body temperatures.

Optimal temperature (°C)

(a)

(b) 40

30

r = 0.76 20 15

25

35

45

(c)

Critical thermal minimum (°C)

Field body temperature (°C)

Critical thermal maximum (°C)

THERMAL SENSITIVITY

65

50

40

r = 0.68 30

25

35

45

Optimal temperature for sprinting (°C)

16 r = 0.08 12

8

4 30

40 Critical thermal maximum (°C)

50

Figure 3.23 Evolutionary correlations were detected between thermal characteristics of 19 species of lizards. (a) The thermal optimum for sprinting evolved in concert with the mean ﬁeld body temperature; the solid reference line has an intercept equal to zero and a slope equal to 1. (b) The critical thermal maximum evolved in concert with the thermal optimum for sprinting. (c) The critical thermal limits evolved independently. Adapted from Huey and Kingsolver (1993, The American Naturalist, University of Chicago Press). © 1993 by The University of Chicago.

Interestingly, the 80% performance breadth was unrelated to variation in ﬁeld body temperature. In general, does the thermal optimum evolve to match the mean temperature experienced by lizards? Compiling data from these and other studies, Huey and Kingsolver (1993) conducted the largest comparative analysis of performance curves for sprinting. Moreover, they used a robust phylogenetic hypothesis to compute independent contrasts for these variables. Contrasts of thermal optima correlated strongly with contrasts of ﬁeld body temperature (Fig. 3.23a), supporting the prediction that the thermal optimum should track the temperature experienced most frequently. If sprint speed were to evolve according to current models, we should also expect corresponding shifts in the CTmin and

CTmax ; however, only the CTmax evolved in concert with the thermal optimum (Fig. 3.23b and c). Thus, species with higher thermal optima can sprint over a wider range of temperatures than species with lower thermal optima. This result disagrees with the notion that specialist–generalist tradeoffs constrain the evolution of this performance curve. Such a result also sharply contrasts the patterns observed in frogs, whose evolutionary history seems to have been constrained by specialist–generalist tradeoffs (see above). As with the thermal sensitivity of survivorship, the thermal sensitivity of locomotion sometimes evolves according to the models described in Section 3.3, but does not always do so. For some taxa, important discrepancies exist between the predicted

66

T H E R M A L A D A P TAT I O N

and observed relationships among thermal heterogeneity, thermal optima, and performance breadths. The agreement between theory and observation will get considerably worse as we consider other performances.

3.4.3 Development From earlier in Section 3.4, recall two predictions about the thermal sensitivity of development. First, the thermal optimum should match the mean (or modal) temperature. Second, the performance breadth should scale according to the relative magnitudes of thermal heterogeneity within and among generations; for short-lived species, the performance breadth should increase with increasing latitude because seasonal variation would occur primarily among generations. Unfortunately, optimality models do little to help us understand latitudinal variation in the thermal sensitivity of development. Although few comparative studies of development have adequately deﬁned the shapes of performance curves, none of these studies suggests these curves have evolved according to Gilchrist’s model (see Table 3.6). A few other studies provide either partial or indirect support for the model. For example, green frogs from a montane population developed faster at low temperatures but slower at high temperatures than did frogs from a lowland population (Berven et al. 1979). Similarly, copepods have diverged along latitudinal clines; genotypes from high latitude developed faster at low temperatures than genotypes from low latitude, and vice versa at high temperatures (Lonsdale and Levinton 1985). More often, investigators have uncovered either no latitudinal pattern or an unexpected pattern. Clones of pea aphids (Acyrthosiphon pisum) from ﬁve populations did not possess very different thermal optima or breadths, and neither parameter was related to latitude or temperature (Lamb et al. 1987). In fact, variation among clones within populations was greater than variation among populations. Hoffmann and colleagues (2005) raised genotypes of the mouse-ear cress (Arabidopsis thaliana) from 65 populations throughout the world. Their results were exactly the opposite of what we should expect

given thermal adaptation; plants from cold environments developed slower at 14◦ C and faster at 22◦ C than plants from warm environments. By comparing performance curves for development along latitudinal and altitudinal transects, investigators have uncovered many examples of countergradient and cogradient variations (Fig. 3.19c and d). Cogradient variation has been documented in common garden studies of insects, including ant lions (Arnett and Gotelli 1999) and fruit ﬂies (Norry et al. 2001). Countergradient variation has been documented in similar studies of reptiles and amphibians. In two species of lizards, embryos from cold environments developed faster under cool and warm cycles of temperature than embryos from warm environments (Ouﬁero and Angilletta 2006; Qualls and Shine 1998). In frogs (Rana spp.), countergradient variation occurs on both macro- and microgeographic scales. Among ﬁve populations from latitudes ranging from 55◦ to 70◦ N, tadpoles at higher latitudes developed faster at constant temperatures (14−22◦ C) (Laugen et al. 2003a; Merila et al. 2000). Along altitudinal transects, tadpoles from mountain ponds developed faster over wide ranges of temperatures than did tadpoles from lowland ponds (Berven 1982; Berven and Gill 1983). On a microgeographic scale, tadpoles from shaded ponds developed faster at two constant temperatures than tadpoles from unshaded pond (Skelly 2004). The abundance of evidence for cogradient and countergradient variations challenges our theory, which assumes a jack of all temperatures will be a master of none. If we take these patterns at face value, we must conclude a genotype can be a master of all temperatures.

3.4.4 Growth Geographic variation in the thermal sensitivity of growth rate has been documented for many species (see Table 3.6). Here, I deﬁne growth broadly to include everything from the change in mass (or length) to the assimilation of energy from the environment (photosynthesis). As with developmental rate, we can explain very few observations using insights from Gilchrist’s model.

THERMAL SENSITIVITY

Growth rate (% inoculum)

800 2

600

1

3 4 5

6 7

400

200 0 16

18

20

22

24

26

28

30

32

Temperature (°C) Figure 3.24 The thermal optimum for the growth of symbiotic algae (Trebouxia irregularis) isolated from lichens of the genus Cladina varied with latitude. The relationship between the temperature of the environment during summer and the thermal optimum was quite strong (r 2 = 0.74, P < 0.005). Source localities: 1, Florida, USA; 2, North Carolina, USA; 3, Rhode Island, USA; 4, Oaxaca, Mexico; 5, Ontario, Canada; 6 and 7, Alaska, USA. Adapted from Sun and Friedmann (2005) with kind permission from Springer Science and Business Media. © Springer-Verlag 2005.

Studies of growth, in terms of biomass or length, have yielded very weak support for the current theory. The best support comes from a study of symbiotic algae (Trebouxia irregularis) isolated from seven populations of lichens (Cladina spp.) along a latitudinal transect (Sun and Friedmann 2005). The performance breadths of the algae were remarkably uniform for all populations, but the thermal optimum correlated strongly with latitude (Fig. 3.24). In other words, algae from cold and warm environments were specialized for low and high temperatures, respectively. Several other studies provide mixed support. Australian trees from temperate rainforests exhibited slightly lower thermal optima than did trees from tropical rainforests. Nevertheless, performance curves for temperate and tropical trees overlapped greatly (Cunningham and Read 2003a). Similar patterns of thermal adaptation were observed in ﬁsh (Belk et al. 2005; Schultz et al. 1996) and frogs (Berven et al. 1979; Stahlberg et al. 2001), but divergences in thermal optima between populations were minimal and genotypes grew submaximally within their natural range of environmental temperatures.

67

Many other researchers have documented patterns that cannot be explained by Gilchrist’s model. In most cases, genotypes from one environment outperform genotypes from other environments over a wide range of temperatures (i.e., countergradient or cogradient variation). For example, water ﬂeas (Daphnia magna) spanning Finland to Sicily exhibited cogradient variation, with no evidence of divergence in the thermal optimum for growth rate among populations (Fig. 3.25c). Countergradient variation has been observed in a wide range of animals, including insects (Van Doorslaer and Stoks 2005a), crustaceans (Lonsdale and Levinton 1985), and ﬁsh (Imsland et al. 2000; Jonassen et al. 2000; Yamahira et al. 2007). The best studied case of countergradient variation occurs within the Atlantic silverside, Menidia menidia (Conover and Present 1990). Genotypes from high latitudes grow as fast as or faster than genotypes from low latitudes. A similar pattern has also been documented within the more narrowly distributed congener, M. peninsulae (Fig. 3.25d). Finally, certain patterns correspond to neither cogradient nor countergradient variation but still cannot be explained by Gilchrist’s model. For example, genotypes of herbaceous plants (Vicia pisiformis and V. dumetorum) from intermediate latitudes exhibited faster growth over a range of temperatures than did genotypes from high and low latitudes (Black-Samuelsson and Andersson 1997). Furthermore, marine algae (Fibrocaspa japonica) from very stable waters off the coast of New Zealand exhibited a wider performance breadth than those of conspeciﬁcs from more variable waters off the coasts of Germany and Japan (De Boer et al. 2005). Thermal sensitivities of photosynthesis also seem difﬁcult to explain with Gilchrist’s model. Although thermal optima and performance breadths vary greatly among plants (reviewed by Berry and Bjorkman 1980; Davison 1991), rigorous comparative studies have generated some unexpected patterns. Cunningham and Read (2002) compared the thermal sensitivities of net photosynthetic rate between temperate and tropical trees. Consistent with the theory, trees from higher latitudes underwent maximal rates of photosynthesis at lower temperatures than trees from lower latitudes. Because a generation for a tree spans many seasons, thermal heterogeneity among seasons should have

T H E R M A L A D A P TAT I O N

(a)

(c) 0.9 FS1 MM SD LB OM CN

DO

Specific somatic growth rate (day–1)

68

0.8

Spain Sicily S. Germany N. Germany

0.7 0.6 0.5

OM FS1 MM SD

0.4 0.3

S2

(b)

DO S2 CN LB

17

20

Belgium Finland Russia England

23 26 Temperature (°C)

(d)

29

NS 5.0 Menidia menidia Menidia peninsulae

NS 40° N

NY Menidia menidia SC SA

30° N

NSB VB Menidia peninsulae

AB

Growth rate (mg day–1)

4.0 NY 3.0

SC SA

2.0

NSB VB

1.0

GK

GK 100° W

90° W

80° W

70° W

AB

60° W

0 15

20

25 30 Temperature (°C)

35

Figure 3.25 Cogradient variation in growth rate has evolved among populations of water ﬂeas, Daphnia magna, while countergradient variations in growth rate have evolved among populations of Atlantic silversides, Menidia menidia and Menidia peninsulae. (a) Sites throughout western Europe where water ﬂeas were collected: Munich, Germany (CN); Doñana, Spain (DO); Storgrundet, Finland (FS1); Lebrade, Germany (LB); Moscow, Russia (MM); Oude Meren, Belgium (OM); Sicily (S2); South Dalton, UK (SD). (b) Sites near the eastern coast of the USA where ﬁsh were collected: Nova Scotia (NS); New York (NY); South Carolina (SC); and St. Augustine (SA), Apalachee Bay (AB), New Smyrna Beach (NSB), Vero Beach (VB), and Grassy Key (GK), Florida. (c and d) The thermal sensitivities of growth, measured in terms of mass. Plots for Daphnia magna were reproduced from Mitchell and Lampert (2000) with permission from Blackwell Publishing. Plots for Menidia spp. were reproduced from Yamahira and Conover (2002) with permission from the Ecological Society of America.

resulted in the evolution of thermal specialists (see Section 3.3.2). Yet trees from more seasonal environments possessed a wider thermal breadth of photosynthesis (Fig. 3.26). Thermal sensitivities of photosynthesis by aquatic macrophytes are even more difﬁcult to explain. Clones of Potamogeton pectinatus, collected over latitudes of 42◦ to 68◦ , possessed very similar performance curves (Pilon and Santamaria 2002). Worse still, species of aquatic macrophytes collected from high latitudes had

higher thermal optima than did species collected from low latitudes—the opposite of what should emerge from thermal adaptation (Santamaria and van Vierssen 1997).

3.4.5 Reproduction Relatively few researchers have investigated performance curves for reproduction along geographic

THERMAL SENSITIVITY

16

(b) 80% performance breadth for photosynthesis (°C)

80% performance breadth for photosynthesis (°C)

(a)

14 12 10 8 7

8

9 10 11 Annual range of maximal daily temperatures (°C)

12

69

16

14

12

10

8

8

9 10 11 12 Mean daily range of temperatures (°C)

Figure 3.26 Performance breadth for photosynthesis increased with increasing thermal heterogeneity during the year (a), but was unaffected by the magnitude of thermal heterogeneity during the day (b). Data are from Tables 2 and 3 of Cunningham and Read (2002).

clines (see Table 3.6). The most extensive comparison was made by Grewal and colleagues (1994), who compared the performance breadths of 12 species of parasitic nematodes from different regions of the world. Performance breadths differed between species from the same locality but did not differ among species from different localities. Hence, Grewal concluded thermal adaptation had not occurred. In contrast, recent studies of butterﬂies on a microgeographic scale yielded less surprising results. Butterﬂies from warm agricultural habitats laid more eggs at high temperatures and fewer eggs at low temperatures than did butterﬂies from cool wooded habitats (see Fig. 3.6e); this pattern held both within species (Karlsson and Van Dyck 2005) and between species (Karlsson and Wiklund 2005). Lastly, the thermal limits of fertility differed between males of temperate and tropical strains of D. melanogaster; the CTmin and CTmax of tropical males exceeded those of temperate males (David et al. 2005; Rohmer et al. 2004). Although these studies are too few to draw a general conclusion, Gilchrist’s model explains some, but not all, of these data.

3.4.6 Why do certain patterns differ from predicted ones? The simple models described in Section 3.3 make many assumptions, some of which will be invalid for any species. These assumptions fall into four broad categories: (1) assumptions about the state of

the environment, (2) assumptions about the nature of phenotypic constraints, (3) assumptions about the relationship between the phenotype and ﬁtness, and (4) assumptions about the presence of genetic variation. First, we might have incorrectly related the thermal conditions of natural environments to the conditions of the models, including the mean, mode, or variance of operative environmental temperature. Second, the models might have incorrectly deﬁned have the phenotypic set, including the range of phenotypes that organisms can express (e.g., thermal optima and performance breadths) and the tradeoffs that constrain the covariation between traits (e.g., specialist–generalist tradeoffs). Third, the models might have incorrectly deﬁned the ﬁtness set, including the timing or nature of how performance contributes to ﬁtness. Finally, the hard-core assumptions of optimality models might have been invalid for some of the organisms considered above; from Chapter 1, recall that several assumptions must be made to infer evolutionary equilibria from optimality models, including the existence of genetic variation in performance curves (see Section 1.4.1). When one or more of these assumptions are violated, we can expect real patterns to differ from predicted patterns. Given the mixed support for the predictions of optimality models (Table 3.7), we should expect that some of the assumptions of the theory do not apply very generally to the species that have been studied thus far. By questioning our assumptions, we can learn how to develop more accurate models of the evolution of thermal

70

T H E R M A L A D A P TAT I O N

Table 3.7 A summary of comparative studies of thermal performance curves listed in Tables 3.3 and 3.6. The majority of these studies provided evidence of evolutionary divergence between populations, but only a minority of them provided support for the predictions of optimality theory

Performance

Case studies

Evidence of evolution

Support for theory

Survivorship Locomotion Growth Development Reproduction

8 7 11 4 1

4 6 9 2 0

2 6 5 0 0

sensitivity. Therefore, the remainder of this chapter focuses on the validity of these assumptions and the models that might provide deeper insights.

3.5 Have we mischaracterized thermal clines? To predict geographic variation in performance curves, we assumed some knowledge of the temperatures experienced by organisms. Although clear latitudinal and altitudinal clines in temperature exist (see Chapter 2), temporal and spatial heterogeneity in real environments likely differ from the conditions assumed in current optimality models. Unlike the normally distributed temperatures of Lynch and Gabriel’s model, temporal variation is bimodally distributed. Furthermore, temperatures in natural environments are correlated over time because of seasonal changes in climate. Gilchrist’s model accounts for the bimodal distribution and the temporal autocorrelation of temperature, yet this model also fails to explain many empirical observations. Perhaps the bigger problem is that both models assume that organisms conform to the operative temperature of their environment. Yet thermoregulation decouples the mean temperature of an organism from the mean temperature of its environment. In doing so, thermoregulation can inhibit the evolution of performance curves along geographic clines (Huey et al. 2003). In Chapter 4, we shall take a closer look at the evolution of thermoregulation. Lastly, neither optimality model considers the possibility

that environmental temperature could change directionally over time, such as the recent warming of many regions throughout the planet. In Chapter 9, we shall return to this topic and examine the consequences of continuous warming for the evolution of performance curves. Do these complexities really explain why geographic patterns do not conform to the predictions of optimality models? Or is there something fundamentally wrong with another aspect of the theory? To answer these questions, we can consider the results of two kinds of experiments: reciprocal transplants and laboratory selection (Kassen 2002). In a reciprocal transplant experiment, one moves a subset of individuals from one thermal environment to another thermal environment, and vice versa. If each population has adapted to its respective environment, genotypes will achieve greater ﬁtness in their native environment than they will in a foreign environment. Furthermore, the native population should outcompete the transplanted population in each environment (Kawecki and Ebert 2004). Still, transplant experiments provide only weak support for models of thermal adaptation because environments differ in more ways than temperature. Laboratory selection experiments offer more powerful tests of evolutionary models. In a selection experiment, one exposes replicated populations to controlled temperatures and observes the evolutionary responses (Section 1.5.2). Thus, we can easily infer the selective environment and relate the evolutionary responses to our theoretical predictions. Nevertheless, artiﬁcial environments of the laboratory lack many of the important elements of natural environments (predators, parasites, competitors, etc.), preventing us from generalizing conclusions from laboratory experiments. Because both reciprocal transplant and laboratory selection experiments have distinct advantages and disadvantages, let us examine examples of each to see whether they support the predictions of optimality models.

3.5.1 Reciprocal transplant experiments Reciprocal transplant experiments have revealed strong evidence of adaptation to local environments. Schluter (2000) reviewed more than 40 experiments

THERMAL SENSITIVITY

in which the performance or ﬁtness of native and transplanted genotypes were compared in multiple environments. In all but eight experiments, native genotypes outperformed transplanted genotypes in each environment. Some degree of adaptation to the local environment seems the norm rather than the exception. But to what extent has this adaptation been driven by thermal heterogeneity? If thermal adaptation has occurred, we should expect a correlation between the relative performance of a genotype and the thermal conditions of the environment. Such a correlation was observed when populations of the carline thistle (Carlina vulgaris) were transplanted among ﬁve regions of Europe (Becker et al. 2006). The ﬁtness of plants was greatest in their native region and decreased as the difference in temperature between the native and transplant regions increased. An experiment conducted by Chapin and Chapin (1981) provided weaker support for the current theory, in the sense that only one of two performances followed the expected pattern. These researchers transplanted populations of a sedge (Carex aquatilis) among ﬁve locations, ranging from an Arctic site (4◦ C) to a subalpine site (17◦ C). Only Arctic plants survived in the Arctic, but these same genotypes did not survive in three of the warmer sites. Roughly speaking, growth rates in localities where multiple genotypes survived were consistent with cogradient variation. Other experiments yielded results consistent with countergradient variation or no variation in performance curves. Cleavitt (2004) reciprocally transplanted fragments of a subalpine moss (Mnium arizonicum) and a lowland moss (M. spinulosum). In both environments, the subalpine species survived as well as or better than the lowland species. Furthermore, quantum yield was greater for subalpine adults in both environments. By transplanting tadpoles between lowland and mountain ponds, Berven (1982) observed faster growth (but not faster development) of montane genotypes in both environments. Reciprocal transplants of lizards (Sorci et al. 1996) and frogs (Morrison and Hero 2003a) along altitudinal gradients yielded no evidence of adaptation to local environments. Overall, reciprocal transplant experiments conﬁrm the conclusion we drew from a survey of comparative data: simple optimality models cannot account for some of

71

the variation in thermal sensitivity observed along geographic clines.

3.5.2 Laboratory selection experiments Laboratory experiments enable researchers to directly explore the evolutionary consequences of speciﬁc selective environments (see Chapter 1). To date, laboratory selection of short-lived organisms— viruses, bacteria, and fruit ﬂies—have contributed to our understanding of thermal adaptation. In each experiment, a population was maintained at a single constant temperature for many generations to enable adaptation to the artiﬁcial conditions of the laboratory. Then, the population was subdivided to form selection lines. Each line was exposed to one of several thermal environments for another series of generations. At the end of this period, the performance curves of these lines were compared. For each experiment, optimality models predict speciﬁc outcomes; for example, selection at a constant temperature should produce a thermal specialist that performs maximally at this temperature. Bennett, Lenski, and their colleagues have conducted some of the most elegant experimental studies of thermal adaptation. Taking advantage of Escherichia coli’s short generations, ease of culture, and genetic markers, these investigators designed a selection experiment that meets the rigorous standards required to infer adaptation (see Box 1.4). First, selection lines were established from populations that had been raised under laboratory conditions for 2000 generations; thus, the investigators were unlikely to have confounded adaptation to artiﬁcial conditions with adaptation to environmental temperature. Second, sufﬁcient replication was used to generate conﬁdence intervals for the mean response to selection. Third, genetic markers enabled the researchers to eliminate the possibility that contamination or migration caused the observed patterns. Finally, a remarkable feature of E. coli makes these studies of adaptation truly unique: E. coli survives prolonged exposure to freezing temperatures, enabling Bennett and colleagues to store ancestral populations and directly compare the performances of ancestral and derived lineages.

72

T H E R M A L A D A P TAT I O N

empirical observations are the most impressive evidence of thermal adaptation to date. Still, some important predictions of the theory could not be conﬁrmed. Optimality models assume that adaptation to the mean or modal operative temperature will come at the expense of ﬁtness at other temperatures. This phenomenon was seen in some lines but not others. Adaptation to 20◦ C was accompanied by a decrement in ﬁtness at 42◦ C, but the converse was not true. In fact, most lines maintained their tolerance ranges despite the expectation that specialists would have evolved at constant temperatures (Fig. 3.27). Perhaps the duration of selection was insufﬁcient for specialists to evolve. In a similar experiment, selection for 20 000 generations at 37◦ C resulted in increasing specialization over the ﬁrst 10 000 generations, with signiﬁcant increases in specialization occurring only after the ﬁrst 2000 generations (Cooper et al. 2001). Huey and colleagues (1991) compared the performances of ﬂies that had evolved at 16.5◦ C and 25◦ C for over 4 years. They found only weak support for the predictions of Gilchrist’s model. Flies that evolved at 16.5◦ C developed faster at 16.5◦ C and slower at 25◦ C than ﬂies that evolved at 25◦ C. Nevertheless, differences in developmental rates were small and both sets of selection lines developed faster at 25◦ C than they did at 16.5◦ C. Hence, any shift in the thermal optimum was likely to be very slight. Possibly, the thermal limits of survivorship responded to selection because ﬂies that evolved at

Bennett et al. (1992) used a population of E. coli that had been cultured at 37◦ C for 2000 generations. The population was raised in ﬂasks of medium and was diluted daily through serial transfers. After an initial period, the population was divided into 24 selection lines, each of which was assigned to a thermal environment. Six lines were shifted up by 5◦ C (42◦ C lines), six lines were shifted down by 5◦ C (32◦ C lines), and six lines were cycled daily between 32◦ C and 42◦ C (32−42◦ C lines); a fourth set of six lines remained at 37◦ C to enable the researchers to consider further adaptation to laboratory conditions. These lines were maintained at their respective thermal regime for 2000 generations. After this period of selection, the thermal sensitivities of ﬁtness were described in two ways. First, they estimated absolute ﬁtness in terms of clonal growth. Second, they calculated the ﬁtness of the derived line relative to its ancestor by resuscitating the ancestral genotype and growing it alongside the descendants (Bennett and Lenski 1993, 1996; Bennett et al. 1992). During the experiment, all lines adapted to their thermal environment. In fact, adaptation of the thermal optimum was nearly perfect! Thermal optima for lines that evolved at 32◦ C, 37◦ C, and 42◦ C equaled 30◦ C, 37◦ C, and 42◦ C, respectively. Moreover, lines in the ﬂuctuating environment evolved a thermal optimum equal to the upper mode (42◦ C), which corresponds exactly to the prediction of Gilchrist’s model. In my mind, these matches between theoretical predictions and

1.50 1.25 Fitness relative to ancestor

1.50

32°C lines

*

1.00

1.00

0.75

0.75

0.50 1.50

20

25

30

35

40

0.50 1.50

42°C lines

1.25

*

1.00 0.75 0.50

37°C lines

1.25

*

*

1.25 1.00

*

20

25

30

35

40

32/42°C lines

*

* *

*

0.75 20

25

30

35

0.50 40 20 25 Temperature (°C)

30

35

40

Figure 3.27 Experimental evolution of Escherichia coli at 32◦ C, 37◦ C, 42◦ C, or 32–42◦ C for 2000 generations increased ﬁtness in the selective environment without decreasing ﬁtness in other environments. Plots show the mean ﬁtness of selected lines relative to the mean ﬁtness of the ancestral line over a range of temperatures. Asterisks denote temperatures at which the ﬁtness of the selected lines differed signiﬁcantly from the ﬁtness of the ancestral line. Adapted from Bennett and Lenski (1993) with permission from Blackwell Publishing.

THERMAL SENSITIVITY

(a)

4 G4 at 44°C

18

14 12

2

Fitness 10 Mutations 8

1

6 0 50

70

90 Passage

110

4 130

(b)

6 5

4

φX174 at 44°C

Fitness

4

3 Mutations

3 2 2 1

1

0 –1

0

10

20 30 Passage

40

50

Number of mutations

3

Number of mutations

Fitness of G4 (doublings per generation)

16

increased exponentially during incubation. Following incubation, a subsample of the viral population was transferred to a new ﬂask of naïve bacterial cells and the cycle continued. Large populations (106 ) were used to minimize genetic drift during the experiment. Selective regimes were tailored to each phage. The investigators knew one of the phages (G4) tolerated high temperatures poorly; therefore, they cultured this phage at 37◦ C for 40 transfers, then at 41.5◦ C for 50 transfers, and ﬁnally at 44◦ C for 50 transfers. The other phage (φX174) was cultured directly at 44◦ C for 50 transfers. At certain intervals, the ﬁtnesses (or doubling rates) of the phages were measured at 37◦ C and 44◦ C to assess whether adaptation had occurred. The ﬁtness of the G4 phage increased slowly and steadily throughout the experiment, but the ﬁtness of the φX174 phage increased very sharply within ﬁve generations and remained relatively constant thereafter (Fig. 3.28). During adaptation to 44◦ C, the φX174 phage suffered a slight decrease in ﬁtness at 37◦ C, but the G4 phage suffered no decrease in ﬁtness at this ancestral temperature. Possibly, loss of ﬁtness at lower temperatures occurred but went undocumented. Following up on this possibility, Knies and colleagues (2006) analyzed the pattern of adaptation in the G4 phage using

Fitness of φX174 (doublings per generation)

16.5◦ C were less likely to survive a brief exposure to 39.5◦ C than were ﬂies that evolved at 25◦ C. One year following this study, the lines at 25◦ C were divided to create an additional set of lines, which were shifted to 29◦ C. After 4 additional years of evolution, James and Partridge (1995) remeasured developmental times in the three sets of lines. At temperatures ranging from 16.5◦ C to 29◦ C, the lines that evolved at 16.5◦ C developed faster than the lines that evolved at either 25◦ C or 29◦ C. Countergradient variation, which has been documented often among natural populations (see Section 3.4), had evolved in the laboratory! The survivorship following heat shock still differed predictably among lines. However, the knockdown temperature, the survivorship following cold shock, and the thermal optimum for sprinting differed very slightly or not at all (Gilchrist et al. 1997). Recently, viruses have been added to the family of organisms used to study experimental adaptation to temperature. Holder and Bull (2001) permitted two bacteriophages to evolve through serial transfers at a constant temperature. Each phage was introduced into a culture ﬂask with an excess of bacterial cells and was incubated for a period of time. Each infected bacterial cell lysed to yield approximately 100 viral progeny, such that the viral population

73

0.6 0.4 0.2 0

Figure 3.28 Experimental evolution of bacteriophages at 44◦ C revealed slow and steady adaptation by one phage (a) and rapid adaptation by another (b). Each plot shows the accumulation of mutations (ﬁlled circles) in a phage, along with the change in ﬁtness (open circles). Error bars denote standard errors of the mean of four assays. Adapted from Holder and Bull (2001) with permission. All copyright is retained by the Genetics Society of America.

74

T H E R M A L A D A P TAT I O N

Box 3.4 Template mode of variation: a novel statistical method for describing variation among performance curves Assuming performance curves have a common form or template, one can describe variation in these curves among genotypes in terms of the shifting or stretching of this template. Kingsolver (2001) described three modes of variation: hotter–colder, faster–slower, and specialist–generalist (Fig. 3.29). The hotter–colder and faster–slower modes describe horizontal and vertical shifts, respectively. The specialist–generalist mode describes a simultaneous horizontal stretch and vertical shift. Izem and Kingsolver (2005) showed how one can estimate the percentage of variation associated with each of the modes. First, a template model must be ﬁtted to data for multiple genotypes. These investigators used a simple model with three parameters of interest: 1 1 Pi (Tb ) = ai + P (Tb − mi ) , (3.15) bi bi

(b)

(c)

Performance

(a)

where Pi equals the performance of genotype i , and the parameters ai , mi , and bi set the mean performance, thermal optimum, and performance breadth, respectively, of the template. By examining differences in these parameters among genotypes, one can determine the degree to which each of the three modes contributes to variation in the performance curve. Importantly, the specialist–generalist mode should not be confused with the specialist–generalist tradeoffs discussed throughout this chapter. The specialist–generalist mode deﬁnes a pattern of variation in the performance curve, whereas specialist–generalist tradeoffs deﬁne a class of mechanisms that lead to such patterns. Actually, both the faster–slower and the specialist–generalist modes can result from specialist–generalist tradeoffs. Therefore, the concepts of modes and tradeoffs are related but not equal.

Temperature Figure 3.29 Using template mode of variation, one can partition the variation in a performance curve among three modes: (a) faster–slower, (b) hotter–colder, and (c) specialist–generalist. These modes reﬂect changes in the parameters of a template function. Adapted from Knies et al. (2006) with permission from the Public Library of Science.

a method known as template mode of variation (Box 3.4). They obtained samples of this phage from the ancestral and selected populations and measured the thermal sensitivity of ﬁtness over a much broader range of temperatures (27−44◦ C). A shift in the thermal optimum accounted for almost 50% of the evolutionary change in the thermal reaction norm. This analysis nicely demonstrated that adaptation to high temperature caused maladaptation to low temperatures, as assumed by optimality models.

3.5.3 Conclusions from reciprocal transplant and laboratory selection experiments Reciprocal transplant and laboratory selection experiments have corroborated the failure of optimality models to explain the evolution of thermal sensitivity. Many reciprocal transplant experiments revealed evidence of local adaptation, but adaptation to one environment does not always cause maladaptation to other environments. Selection experiments with E. coli and D. melanogaster demonstrated conclusively that a jack of all temperatures need not

THERMAL SENSITIVITY

be a master of none. Furthermore, the experiment with D. melanogaster showed that thermal optima sometimes evolve slowly even in the face of strong selective pressure. Given the designs of these experiments, we cannot attribute the mismatch between theory and observation to poor knowledge of the thermal environment. Instead, we must question other assumptions of optimality models.

3.6 Have phenotypic constraints been correctly identiﬁed? Two constraints deﬁne the phenotypic functions of the optimality models described in this chapter: (1) performance curves cannot change by acclimation and (2) performance curves must maintain a constant area during evolution. The ﬁrst constraint prevents the hypothetical organism from adjusting its thermal sensitivity of performance during development. In Chapter 5, we shall consider the potential for thermal acclimation and examine models that relax this constraint. The second constraint prevents an organism from adapting to one temperature without losing function at another temperature. Because specialist–generalist tradeoffs have been a major focus of this chapter, let us now consider the empirical evidence for this constraint.

3.6.1 A jack of all temperatures can be a master of all Much evidence stands against the idea that specialist–generalist tradeoffs universally constrain performance curves. Numerous examples can be found among the comparative studies discussed in Section 3.6. Even the earliest comparisons of performance curves indicated that a jack of all temperatures could also be a master of all; for example, Huey and Hertz (1984) found that the rank order of performance among individuals of several species was fairly conserved across a range of temperatures (ca. 20◦ C). Much experimental work, including both quantitative genetic and artiﬁcial selection experiments, now corroborates this comparative evidence. A comparison of wasp strains raised under common environmental conditions revealed little or no correlation between the performance breadth and the maximal performance (Carriere and Boivin

75

1997), whereas a comparison of aphid families revealed a negative genetic correlation for males but no genetic correlation for females (Gilchrist 1996). Among 29 genotypes of Daphnia pulicaria, no relationship existed between the thermal breadth of ﬁtness (r) and the maximal ﬁtness (Palaima and Spitze 2004). Using template mode of variation (Box 3.4), Izem and Kingsolver (2005) discovered that only half of the variation among families of butterﬂies reﬂected some form of specialist–generalist tradeoff; the remaining half included variation in the height of the performance curve (13%) and variation that could not be explained by their model (33%). Selection experiments have also provided mixed support for the importance of specialist–generalist tradeoffs. Selection for greater performance at one temperature does not always cause a correlated decrease in performance at other temperatures (Anderson et al. 2005; Bennett and Lenski 1993; Carrière and Boivin 2001). As concluded from the survey of comparative data (Section 3.4), a jack of all temperatures does not have to be a master of none.

3.6.2 The proximate basis of performance determines tradeoffs Recall that specialist–generalist tradeoffs arise from the thermal adaptation of enzymes, membranes, and organelles (see Section 3.2). Despite the sound biochemical basis for specialist–generalist tradeoffs, species can apparently avoid these tradeoffs during evolution. How do some genotypes become masters of all temperatures? And what other kinds of tradeoffs prevent these genotypes from taking over the world? When performance curves differ between genotypes, the proximate mechanisms that generate this difference can be quite complex. These mechanisms not only include changes in the structures of enzymes, but can also include changes in concentrations (Somero et al. 1996). Furthermore, enzymes do not function in a vacuum; they interact intimately with their surroundings and these interactions affect the thermal sensitivity of enzymatic function. Additional stability emerges from interactions between enzymes and other molecules, such as ions, substrates, coenzymes, biogenic amines, and ATPconsuming chaperones (Fields 2001; Hochochka and

76

T H E R M A L A D A P TAT I O N

Somero 2002; Jaenicke 1991; Somero 2003). Greater concentrations of paralogous isozymes or extrinsic stabilizers could expand the thermal breadth of performance. Finally, the number, size, or quality of organelles, cells, tissues, and organs should also inﬂuence performance. But any evolutionary increase in performance comes at a price! Enzymes and their stabilizers cost energy to produce or sequester. Organelles, cells, and tissues cost energy to produce and maintain. The energy expended on such adaptations to temperature must be diverted from other functions. When these kinds of thermal adaptation occur, an organism must acquire additional energy to continue performing multiple functions or reallocate energy among competing functions. For instance, an individual that maintains higher concentrations of enzymes could enhance performance at extreme temperatures without a loss of performance at moderate temperatures, but would have less energy to devote to other cellular processes (an allocation tradeoff). If we assume the same individual forages more to offset this energy demand, we should expect this behavior to impose a greater risk of predation or parasitism (an acquisition tradeoff). Therefore, every modiﬁcation of a performance curve results in some type of tradeoff (Box 3.5). To appreciate the importance of these tradeoffs, consider some of the proximate mechanisms for variation in the thermal sensitivity of growth rate. In many cases, populations in colder environments have evolved faster growth at relatively low temperatures (see Table 3.6). Yet the mechanisms that underpin these adaptations can be quite diverse. Thermal adaptation of enzymes decreases the concentrations necessary for anabolic and catabolic processes, leaving more energy available for somatic growth. Growth efﬁciency also increases when organisms divert energy from cellular maintenance, such as protein turnover or ion transport (Bayne 2004; Hawkins 1991, 1995; Hulbert and Else 2000). Finally, intense rates of foraging and consumption fuel rapid growth. Accordingly, species have taken multiple routes to the same evolutionary endpoint. In some species, rapidly growing genotypes allocate more energy to growth and less to maintenance (Imsland et al. 2000;

Lindgren and Laurila 2005; Ouﬁero and Angilletta 2006; Robinson and Partridge 2001). In other species, genotypes achieve rapid growth through high rates of energy intake (Jonassen et al. 2000; Nicieza et al. 1994). The evolution of growth rate often involves simultaneous changes in food consumption and growth efﬁciency (Billerbeck et al. 2000; Jonsson et al. 2001; Lonsdale and Levinton 1989; Stoks et al. 2005; Van Doorslaer and Stoks 2005a). Artiﬁcial selection has underscored the interactive role of behavior and physiology in the evolution of growth rate. Aquacultural breeding programs have produced slow- and fast-growing lines of commercially important species. In two species of oysters, rapidly growing genotypes had higher rates of food consumption, lower rates of maintenance metabolism, and lower costs of somatic growth (Bayne 1999, 2000). Tradeoffs arising from an increase in feeding or a reduction in maintenance can affect the ﬁtness of these genotypes in certain environments (Box 3.6). Because a diversity of proximate mechanisms can transform the genotype into the phenotype, we cannot assume all evolutionary changes will give rise to specialist–generalist tradeoffs. Once we begin to consider the interactions among acquisition, allocation, and specialist–generalist tradeoffs, we have entered the realm of life-history theory. This body of theory describes the way in which strategies of acquisition and allocation evolve. In Chapters 6 and 7, we shall see how life-history theory can help us to understand the evolution of performance curves.

3.7 Do all performances affect ﬁtness? At ﬁrst, asking whether performances such as survivorship and reproduction affect ﬁtness sounds silly. Still, optimality models assume a very speciﬁc relationship between performance and ﬁtness. To ﬁnd the optimal performance curve, both Lynch and Gabriel (1987) and Gilchrist (1995) assumed ﬁtness was directly related to performance. In other words, these models propose a form of directional selection characterized by a positive linear selection gradient (see Fig. 1.6a). Does directional selection really approximate the true nature of selection? Perhaps

THERMAL SENSITIVITY

77

Box 3.5 Tradeoffs associated with the evolution of thermal performance curves Recently, my colleagues and I summarized the kinds of tradeoffs that constrain the evolution of performance curves (Angilletta et al. 2003). These tradeoffs are best illustrated by an example: imagine performance curves for growth and reproduction. Suppose a mutation enables a genotype to grow more rapidly at low temperatures. The mutant could grow rapidly because it allocates more energy to growth than the ancestor would; however, the mutant’s increase in growth rate would decrease its reproductive rate (Fig. 3.30a). Life historians refer to such relationships as allocation tradeoffs (Zera and Harshman 2001). Alternatively, the mutant could grow more rapidly by acquiring more energy, such that no energy is diverted from reproduction (Fig. 3.30b). In this case, the mutant could enhance growth rate and reproductive rate simultaneously (Reznick et al. 2000). This mutant will likely experience an acquisition tradeoff, in which an increase in the duration or intensity of feeding increases the probability of death by predation (Gotthard 2000; McPeek 2004; Stoks et al. 2005). Finally, the mutant could grow more rapidly by

X Y X Y

Temperature

(c) Phenotype

(b) Phenotype

Phenotype

(a)

assimilating and using resources more efﬁciently (Wieser 1994). This form of adaptation would require changes in enzymatic structure that impose a specialist–generalist tradeoff (Fig. 3.30c). Importantly, genotypes can combine these different mechanisms to further enhance performance, but multiple tradeoffs would occur. When a performance curve evolves, both the derivative and the integral of the function can change (i.e., the shape and the height can evolve). Mechanisms associated with specialist–generalist tradeoffs inﬂuence the derivative of the curve without changing its integral (see Fig. 3.30c), whereas mechanisms associated with allocation or acquisition tradeoffs inﬂuence the integral of the curve. Because a performance curve results from many interacting mechanisms, we cannot always understand tradeoffs by comparing the shapes of curves among genotypes. Evolutionary thermal biologists must closely examine the proximate mechanisms by which genotypes express different performance curves.

X Y X Y

Temperature

Temperature

Figure 3.30 Three classes of tradeoffs associated with variation in thermal performance curves. Each plot depicts two or more performance curves; the pies depict the quantity of resources allocated to one function (X ) versus another function (Y ) at a given temperature. (a) Given a ﬁxed quantity of resources, a difference in the allocation of resources between X and Y (e.g., growth and reproduction) can cause a difference in performance at all temperatures. An evolutionary change in performance resulting from function X (e.g., growth rate) would be negatively correlated with an evolutionary change in a performance resulting from function Y (e.g., fecundity). (b) A difference in the acquisition of resources can cause a difference in performance at all temperatures. In this case, an evolutionary change in the performance resulting from function X could be positively correlated with an evolutionary change in a performance resulting from function Y. (c) Performance at certain temperatures can be enhanced by thermal specialization, as a genotype can perform more efﬁciently within a narrow range of temperatures or less efﬁciently within a broader range of temperatures. Such specialist–generalist tradeoffs are commonly imposed by assuming the area under the performance curve remains constant (see Section 3.3). Adapted from Angilletta et al. (2003) with permission from Elsevier. © 2003 Elsevier Ltd.

under certain conditions submaximal performance would lead to a greater expectation of ﬁtness, even if it reduces survivorship or reproduction during the short term. Consider an even more outrageous idea: variation in certain performances might not contribute to ﬁtness at all.

To evaluate these ideas, we must quantify selection gradients. Unfortunately, few people have successfully measured selection gradients for performance (Kingsolver et al. 2001b). The reason is simple: researchers often use performance (e.g., survivorship) as a proxy for ﬁtness. This practice

78

T H E R M A L A D A P TAT I O N

Box 3.6 Evolutionary tradeoffs: a case study of ﬁsh variation in the performance curve depend on the phylogenetic scope of analysis. Although variation among populations of M. menidia was caused by different strategies of acquisition and allocation, variation between M. menidia and its southern congener, M. peninsulae, was driven partially by thermal specialization (Yamahira and Conover 2002). Maximal burst speed (cm s–1)

The investigations of countergradient variation in the growth rate of the Atlantic silverside (Menidia menidia) constitute one of the best case studies illustrating the tradeoffs associated with the evolution of performance curves. At all temperatures, ﬁsh from Nova Scotia grow faster than ﬁsh from South Carolina because they consume more food and convert a greater fraction of their food to body mass (Billerbeck et al. 2000; Present and Conover 1992). Presumably, the costs associated with the evolution of growth rate can be reduced by combing behavioral and physiological strategies. Furthermore, these strategies enable northern ﬁsh to grow relatively quickly over a wide range of temperatures. Still, Conover and colleagues presented compelling evidence of a tradeoff between growth rate and swimming speed (Fig. 3.31). When starved, the rapidly growing ﬁsh from Nova Scotia were unable to swim as fast as the slowly growing ﬁsh from South Carolina. Feeding further increased the difference in swimming speed between the two groups. As a consequence, ﬁsh from Nova Scotia suffered a greater risk of predation in staged encounters with predators than ﬁsh from South Carolina (Fig. 3.32). In addition to the obvious tradeoff between feeding and survivorship, the intrinsically slower swimming of northern ﬁsh might have resulted from a poor quality of muscle caused by allocation tradeoffs. Interestingly, the mechanisms that underlie

200 175 150 125 100 NS SC Population and feeding status

Figure 3.31 Atlantic silversides from Nova Scotia (NS) were unable to swim as fast as ﬁsh from South Carolina (SC), regardless of whether they were fed (ﬁlled boxes) or fasted (open boxes). Reproduced from Billerbeck et al. (2001) with permission from Blackwell Publishing.

1.0

Survivorship

0.8 0.6 0.4 0.2 0.0 0

10

20

30 Time (min)

40

50

assumes the very thing we are trying to determine— namely, that ﬁtness depends directly on performance. In reptiles, proxies for ﬁtness only sometimes correlate with locomotor performance

60

Figure 3.32 Survivorship of Atlantic silversides from South Carolina (dashed line) exceeded that of ﬁsh from Nova Scotia (solid line) during staged encounters with predators. Reproduced from Lankford et al. (2001) with permission from Blackwell Publishing.

(reviewed by Angilletta et al. 2002a). We know virtually nothing about selection gradients for other performances, let alone selection gradients for their thermal sensitivity.

THERMAL SENSITIVITY

In fact, despite the hundreds of thermal sensitivities reported in the literature, only one group of researchers has ever estimated a selection gradient for a performance curve. Kingsolver and colleagues (2007) quantiﬁed performance curves for relative growth rate in caterpillars (Pieris rapae). Because these caterpillars grow incredibly fast during late instars, these investigators were able to record growth by the same individuals over a range of temperatures. These acute thermal sensitivities differed from chronic thermal sensitivities (Kingsolver and Woods 1998) and better reﬂect the temporal scale of thermal variation in nature (Kingsolver 2000). After determining the performance curve for each individual, they exposed individuals to one of two thermal regimes: a cool cycle and a warm cycle (Fig. 3.33a). Fitness was estimated as the product of developmental rate, adult survivorship, and adult mass. Thus, the authors were able to quantitatively evaluate the assumption of directional selection. To better appreciate the results of this experiment, let us brieﬂy consider how selection should act on each region of the performance curve. Kingsolver and Gomulkiewicz (2003) proved that, under certain conditions, the strength of selection should depend directly on the frequency with which individuals experience a particular temperature. Therefore, selection gradients for the caterpillars should have been greatest at 13◦ C and 19◦ C in the cool and warm environments, respectively (Fig. 3.33a). In stark contrast to this expectation, selection gradients were maximal at the same temperature in both environments (Fig. 3.33b and c). Why did the predicted selection gradients poorly match the observed selection gradients? Honestly, I do not know. But several considerations regarding the effect of performance on ﬁtness seem worthwhile. First, performance could inﬂuence both survivorship and fecundity in complex ways. If so, the relationship between performance and ﬁtness might depend on the ontogenetic stage of the organism or environmental variables other than temperature. Second, temporal autocorrelation of body temperatures would affect the optimal performance curve when the time-course of performance matters. Third, demographic stochasticity generally leads to a realized ﬁtness that differs from the expected ﬁtness (Burger and Lynch

79

1995). Consequently, simulation models of evolution provide valuable complements to optimality models.

3.8 Does genetic variation constrain thermal adaptation? When using optimality models, we must assume populations in different environments reside at their respective ﬁtness maxima. This primary assumption envelopes two secondary assumptions. First, each population must contain sufﬁcient additive genetic variance to respond to selection. Second, thermal environments must remain fairly consistent over time for each population to reach its evolutionary equilibrium. Either low genetic variance or relatively weak selection could prevent the evolution of optimal genotypes. Additionally, gene ﬂow among populations in different thermal environments could prevent each population from adapting to local conditions. We can explore the consequences of relaxing these core assumptions through genetic models of evolution. Quantitative genetic and allelic models explicitly consider the genetic constraints on adaptation. In Chapter 1, we discussed several ways one could model the genetics of a performance curve (see Section 1.4.2). One way would be to focus on the genetic variance of performance at each of many temperatures. Another way would be to focus on the genetic variance of parameters that describe the performance curve, such as the thermal optimum, performance breadth, and thermal limits. Finally, we could treat the entire performance curve as an inﬁnite dimensional trait and estimate its genetic variance–covariance function. Each of these approaches uniquely deﬁnes the traits under consideration. Because performance curves are continuous functions, researchers have generally adopted the second or third approach. Regardless of how one chooses to model genetics, the evolution of a performance curve depends directly on genetic variances and covariances (Arnold 1992, 1994). The optimal performance curve will evolve eventually, as long as some genetic variance persists and genetic covariances are small enough to permit independent evolution of the performance curve (Gomulkiewicz and Kirkpatrick

80

T H E R M A L A D A P TAT I O N

(a)

Fraction of total

0.12

Cool Warm

0.09

0.06

0.03

0 5

10

15

20

25

30

35

40

(b)

Selection gradient

50 Warm Cool

40 30 20 10 0 –10 10

15

20

25

30

35

(c)

Selection gradient

40

Warm Cool

30 20 10 0 –10 10

15

20

25

30

Temperature (°C)

1992; Scheiner 1993; Via and Lande 1985). Nevertheless, the trajectory of evolution can be circuitous and hundreds of generations might be required to reach an evolutionary equilibrium (Via 1987). Therefore, genetic models provide valuable insights

35

Figure 3.33 Selection gradients for growth rate at different temperatures did not match the predictions of theory. Selection gradients were calculated for caterpillars (Pieris rapae) in cool and warm environments (a) during two separate experiments (b and c). To compute selection gradients, ﬁtness was estimated as the product of survivorship, developmental rate and adult mass. Reproduced from Kingsolver (2007, The American Naturalist, University of Chicago Press). © 2007 by The University of Chicago.

about the dynamics of adaptation (Gomulkiewicz and Kirkpatrick 1992; Via 1994). In this section, we shall examine both quantitative genetic and allelic models of evolution (see Chapter 1). Quantitative genetic models combine

THERMAL SENSITIVITY

selection gradients with genetic variances and covariances to project evolutionary changes. Allelic models consider the impact of mutation, drift, and selection on the course of evolution. Such models were constructed by extending the optimality models of Lynch and Gabriel (1987) and Gilchrist (1995) to include a genetic basis for the performance curve. To apply either kind of model, one must replace the assumption that populations reside at their evolutionary equilibria with assumptions about genes, mutations, and drift. Given these new assumptions, one can simulate the evolution of a performance curve. Below, we shall draw on such simulations to learn how genetic constraints inﬂuence the dynamics of adaptation.

3.8.1 A quantitative genetic model based on multivariate selection theory In a landmark paper, Kingsolver and colleagues (2001a) showed how one can predict the evolution of a performance curve by simply combining knowledge of selection gradients and genetic constraints. These researchers treated a performance curve as an inﬁnitely dimensional trait (see Section 1.4.2). Rather than focus on a discrete matrix of genetic variances and covariances, they focused on a continuous function: G(T, θ ),

(3.16)

which describes the genetic variance of performance at any temperature T and the genetic covariance between any pair of temperatures T and θ. Figure 3.34 shows a genetic variance–covariance function estimated from variation in growth rate among full-sibling families of butterﬂies (Pieris rapae). Correspondingly, a function of selection gradients, β(θ), describes the linear selection gradients over an inﬁnite set of temperatures. Figure 3.35 shows three selection gradient functions for larval growth rate in the same species. The response to selection depends on genetic variances and covariances. The direct effect of selection results from the selection gradient and the genetic variance at T. Indirect effects of selection result from selection gradients at temperatures other than T and genetic covariances between T and θ . These genetic covariances could reﬂect a combination of

81

several different biological processes (see Box 3.5). By integrating the product of the genetic variance– covariance function and the selection gradient function over all temperatures, ¯ P(T) =

G(T, θ)β(θ)dθ,

(3.17)

Kingsolver and colleagues summed the direct and indirect effects of selection on performance at temperature T (Fig. 3.36). Many assumptions must be made to implement this quantitative genetic approach. To estimate the genetic variance–covariance function, we must assume that phenotypic variation among families excludes the effects of dominance, epistasis, and shared environments (see Chapter 1). To estimate the selection-gradient function, we must deﬁne ﬁtness appropriately and quantify it accurately. Furthermore, both functions must remain constant during the period in which we wish to explore the evolutionary dynamics. If either function changes appreciably, the dynamics of evolution will also change. Indeed, we should expect both functions to change during evolution. The genetic variance–covariance function changes systematically under strong selection and genetic drift because mutation and gene ﬂow might not be sufﬁcient to maintain genetic variation (Steppan et al. 2002). Likewise, the selection gradient function will change when the ﬁtness landscape possesses maxima and minima (i.e., peaks and valleys); in such cases, both linear (directional) and nonlinear (stabilizing) selection gradients must be estimated over a suitable range of performance. In the simple example described by Kingsolver and colleagues, only directional selection was considered. Given the assumptions required, this quantitative genetic approach would be most useful for predicting the evolution of performance curves over a few generations.

3.8.2 A genetic model for survivorship and related performances Gabriel (1988) extended the optimality model of Lynch and Gabriel (1987) to include genetic constraints on the evolution of the thermal sensitivity of survivorship. The environmental, phenotypic, and

82

T H E R M A L A D A P TAT I O N

Covariance [(mg h–1)2]

0.4 0.2 0 –0.2 8 11

17 23 Tem 29 pera ture (°C)

35

40

8 11

0.5

Selection gradient

23

°C)

re (

atu

er emp

40

T

Pupal mass Development time Survival

**

0.3

17

35

29

0.1 –0.1 –0.3 –0.5

10

Figure 3.34 A genetic variance–covariance function deﬁnes the genetic constraints on the evolution of a performance curve. This function was estimated from variation in the thermal sensitivities of larval growth rate among families of butterﬂies, Pieris rapae. Reproduced from Kingsolver et al. (2004) with permission from Blackwell Publishing.

15

20 25 Temperature (°C)

30

ﬁtness functions followed Lynch and Gabriel (1987), with the exception that genetic values of the performance curve were modeled via quantitative genetics. Recall from Section 3.3.1 that the performance curve of Lynch and Gabriel’s model was deﬁned by the genotypic values g1 and g2 . In Gabriel’s model, each of these values was deﬁned by the allelic contribution of 20 diploid loci. The environment and the performance curve determined survival to reproduction. Surviving individuals were randomly culled to 250 individuals and these survivors produced the next generation. Reproduction could occur through parthenogenesis or sex, depending on the simulation. Mutation of alleles occurred with a probability of 0.1 per genome.

35

Figure 3.35 A selection-gradient function deﬁnes the strength of selection for performance at each temperature. These functions were estimated for the larval growth rate of a butterﬂy, Pieris rapae. To calculate selection gradients, pupal mass, developmental time, and survival were used as three proxies for ﬁtness. The selection gradient at 11◦ C, estimated from pupal mass, differed signiﬁcantly from zero (as indicated by asterisks). Adapted from Kingsolver et al. (2001a) with kind permission from Springer Science and Business Media. © 2001 Kluwer Academic Publishers.

Using this model, Gabriel asked how the performance curve would respond to a major shift in the mean or variance of operative environmental temperature. For example, if a drastic reduction in the mean temperature occurs, we might expect the thermal optimum (g1 ) to evolve to match the new mean. Although our expectation was conﬁrmed through simulations, the population required hundreds of generations to attain the new optimal value of g1 (Fig. 3.37a). During adaptation, a greater performance breadth (g2 ) evolved to enhance performance of the genotype while the thermal optimum approached the mean environmental temperature (Fig. 3.37b). Once the optimal value of g1 had evolved, the performance breadth returned to its

THERMAL SENSITIVITY

83

(a) 5 0 ∆z–

15

20

25

30

35

–5

–10 –15

– z

(b)

Relative growth rate

25 20 – z′

15 10 5 0

15

20

25

30

Temperature (°C)

original value. Now consider a drastic reduction in the variance of temperature among generations. In this case, we should expect the value of g2 to decrease correspondingly. Again, the process of adaptation required hundreds of generations. Not surprisingly, adaptation in both scenarios proceeded more rapidly when sexual reproduction occurred more frequently.

3.8.3 A genetic model for fecundity and related performances Gilchrist (2000) also extended his optimality model to consider the consequences of mutation, selection, and drift for the evolution of performance curves. Instead of deﬁning distinct loci with allelic contributions to performance, he considered genotypic values for parameters of the performance curve (CTmax and B; see Section 3.3.2). The parameters of

35

Figure 3.36 Predicted evolutionary response of the performance curve for larval growth rate in a butterﬂy, Pieris rapae. (a) Estimates of the selection-gradient function (Fig. 3.35) and the genetic variance–covariance function (Fig. 3.34) were used to calculate the expected change in the performance curve. (b) The predicted performance curve after selection (solid line) differs markedly from the performance curve before selection (dashed line). Adapted from Kingsolver et al. (2001a) with kind permission from Springer Science and Business Media. © 2001 Kluwer Academic Publishers.

the performance curve were ultimately determined by genotypic and environmental values, the latter being normally distributed with an expected value of zero. The genotypic values changed by some quantity during mutation. Gilchrist allowed replicate populations to evolve for 20 000 generations under each of the nine environmental states depicted in Fig. 3.17. Under the combined processes of selection, mutation, and drift, adaptation required many, many generations. In fact, Gilchrist observed a more pronounced lag than we might have expected from Gabriel’s simulations (Fig. 3.38). In certain environments, the performance breadth was still suboptimal after 20 000 generations! Adaptation occurred most slowly when temperature varied greatly within generations because specialists achieved only slightly greater ﬁtness than generalists (see Fig. 3.18). Other factors also contributed to the rate

84

T H E R M A L A D A P TAT I O N

(a)

3.8.4 Predictions of quantitative genetic models depend on genetic parameters

20

8

g1

nsex 10

500 50 ≤ 20 200 400 600 800

(b) 120

g2

8

500 80

50 ≤ 20

40 200 400 600 800 Generations

Figure 3.37 Following an abrupt change in the mean environmental temperature, the thermal optimum (g1 ) and the performance breadth (g2 ) took hundreds of generations to approach their new optima. Various frequencies of sexual reproduction (nsex ) were explored, ranging from at least once per 20 generations (≤ 20) to never (∞). Adapted from Gabriel (1988) with kind permission from Springer Science and Business Media. © Springer-Verlag 1988.

of adaptation. In all environments, a genetic correlation between CTmax and B prevented populations from directly ascending the ﬁtness landscape. This correlation resulted from two assumptions of the model: 1. the performance curve was an asymmetric function (eqn 3.12) rather than a symmetric function as in Lynch and Gabriel’s model (eqn 3.5); 2. the performance curve evolved because of genes for thermal plasticity rather than genes for performance at different temperatures (i.e., the parameter-based approach was used instead of the character-based approach; see Chapter 1). If these assumptions were relaxed, adaptation would have occurred more rapidly (Gilchrist 2000). The ﬁrst assumption seems appropriate for most types of performances, but the second assumption has stimulated controversy (Via et al. 1995).

Quantitative genetic models tell us that adaptation depends on the genetic variance and the genetic architecture of thermal sensitivity. So what do we know about these things in real populations? In most populations examined to date, some degree of genetic variance in thermal sensitivity has been found. Genetic variances have been estimated from similarities among relatives and responses to selection (see Chapter 1). Most studies have focused on parameters of the performance curve such as performance breadth or thermal limits (including indices of thermal tolerance, such as knockdown temperature). Gilchrist (1996) used a breeding experiment to estimate genetic variation in the performance breadth of male aphids (h2 = 0.19, P = 0.06), but could not detect genetic variation among female aphids. The same population possessed no genetic variation in the thermal optimum (h2 = − 0.08, P = 0.32). Similar experiments were used to detect genetic variation in the thermal sensitivity of growth rate in populations of caterpillars (Kingsolver et al. 2004) and ﬁsh (Yamahira et al. 2007). Selection experiments have revealed substantial genetic variance of thermal tolerance in populations of Drosophila and other insects (Anderson et al. 2005; Bubli et al. 1998; Carriere and Boivin 2001; Folk et al. 2006; Gilchrist and Huey 1999; Loeschcke and Krebs 1996). In contrast, Baer and Travis (2000) observed no response to artiﬁcial selection of the critical thermal maximum in a poeciliid ﬁsh (Heterandria formosa). Given that most experiments have detected some degree of genetic variation, one might be tempted to conclude that genetic constraints on adaptation are relatively unimportant. However, two major caveats must be considered. First, the observed genetic variances and covariances do not permit all kinds of evolutionary responses. For example, variation among families of caterpillars suggests the performance breadth can evolve more readily than the thermal optimum (Izem and Kingsolver 2005). Yet, genetic variation among families of ﬁsh suggests the mean performance can evolve more readily than either the thermal optimum or the performance breadth (Yamahira et al. 2007). Second, genetic variation

THERMAL SENSITIVITY

25 20 15 10 5 0

0

5

Variation within generations Medium (a) (b)

High (c)

(d)

(e)

(f)

(g)

(h)

(i)

Variation among generations Low Medium High

Thermal breadth (°C)

Low

85

10 15 20 Generations (× 1000)

Figure 3.38 Simulated evolution of thermal breadth in the nine environments shown in Fig. 3.17. Error bars are standard deviations of the mean response for ﬁve replicate populations. The dashed line marks the optimal breadth. Adapted from Gilchrist (2000) with permission from Elsevier. © 2000 Elsevier Ltd.

in the performance curve depends on the environmental context (Kingsolver et al. 2006). Third, and most importantly, these studies focus on a tiny set of species. Clearly, much more research will be necessary to draw general conclusions about the role of genetic constraints in the evolution of thermal sensitivity. As natural selection and genetic drift proceed, mutation must replenish the genetic variance required for adaptation. Theory indicates genetic variances and covariances for some traits will change markedly over time (Gilchrist et al. 2000; Jones et al. 2003). But how many mutations must occur to sustain thermal adaptation? How many loci contribute to the shape of a performance curve? Are differences in thermal sensitivity caused by many mutations with small effects or a few mutations with large effects? These questions become critical as we try to understand how genetic variation persists during adaptation. Studies of viruses have begun to shed some light on these issues. Apparently, two or three mutations accounted for the majority of adaptation to the extreme temperature of 44◦ C (Holder and Bull 2001). Interestingly, the ﬁtness beneﬁt of these mutations depended on the genetic background; a mutation improved ﬁtness

most when the ﬁtness of the background genotype was lowest (Bull et al. 2000). Many more studies of this kind will be required to characterize the nature of genetic constraints on thermal adaptation.

3.9 Does gene ﬂow constrain thermal adaptation? Until now, we have implicitly assumed that natural selection, mutation, and genetic drift occur in a thoroughly mixed gene pool. Even in Lynch and Gabriel’s model, in which offspring dispersed throughout a spatially heterogeneous landscape, all individuals experienced the same degree of spatial and temporal heterogeneity. But what if a population was more like a set of spatially and genetically structured subpopulations? How would gene ﬂow between subpopulations inﬂuence the evolution of performance curves? Gene ﬂow can prevent populations from adapting perfectly to their environments (Ronce and Kirkpatrick 2001). In the context of thermal physiology, two distinct phenomena can emerge. Thermal optima for performance might diverge between subpopulations, but a balance between selection and

T H E R M A L A D A P TAT I O N

migration will maintain some degree of maladaptation to each environment. Alternatively, similar thermal optima might exist between subpopulations, such that only one of the subpopulations adapts well to its local environment. The latter phenomenon, referred to as migrational meltdown, occurs when migration between populations differs dramatically (Ronce and Kirkpatrick 2001). How does migration affect the evolution of thermal physiology when a species colonizes a novel environment? To address this question, we can draw on recent models by Holt and colleagues (2003, 2004). These investigators viewed the novel environment as a sink because a population of well-adapted genotypes would become maladapted following dispersal from the current range. For simplicity, they assumed a “black-hole” sink, to which some individuals emigrated each generation but from which no individuals ever returned. Their model described a generic trait whose value contributed to ﬁtness (w) via survivorship:

w = Pmax exp

−

(¯g−θ)2 2(σ 2 +VP )

For a given level of maladaptation, the population would be more likely to adapt when more individuals immigrated per generation (Fig. 3.39a). A wide performance breadth or temporal variation in temperature would also promote adaptation. Moderate stochastic variation in temperature promotes adaptation, but extreme stochastic variation inhibits adaptation because certain periods cause severe maladaptation (Fig 3.39b). When temperatures are temporally autocorrelated or cycle periodically, the population adapts more readily because longer stretches of good periods enable the population to rebound. This result has interesting implications for

1.0

(a) 16

0.8 4 8 0.6

I=2

0.4

,

(3.18)

where Pmax equals the survivorship of a perfectly adapted genotype, g¯ equals the mean genotypic value, θ equals the optimal phenotype in the sink, VP equals the phenotypic variance, and 1/σ 2 determines the strength of selection. For our purposes, we can think of this trait as a symmetrical performance curve with a thermal optimum equal to g and a performance breadth proportional to σ 2 . Holt and colleagues explored the process of adaptation with continuous analytical models and individual-based simulations; the latter approach also included stochastic demographic processes resulting in genetic drift and possibly extinction. The probability of adaptation depends on many factors, including the degree of initial maladaptation, the size of the population, the rate of immigration, the variation in temperature, and the thermal performance breadth. Under any set of conditions, the population would be more likely to adapt when its mean thermal optimum more closely corresponds to the mean operative temperature of the environment. A critical level of maladaptation exists, above which adaptation never occurs because the population shrinks severely during selection.

Probability of adaptation

86

0.2 0.0 2.50

2.75

3.00

3.25

3.50

1.0 0.8

(b)

0.6 0.4 0.2 0.0

Coefficient of variation 0 0.1 0.2 0.4 2.6

3.0 3.2 2.8 Mean maladaptation

3.4

Figure 3.39 The rate of migration and the variation in temperature affect the probability of adaptation to a novel environment. (a) For any degree of maladaptation, the probability of adaptation increases as the number of immigrants per generation (I) increases. (b) For any degree of maladaptation, the population would be more likely to adapt when environmental temperature varies moderately over time. (a) Adapted from Holt et al. (2004) with permission from the Royal Society. (b) Adapted from Holt et al (2003, The American Naturalist, University of Chicago Press). © 2003 by The University of Chicago.

THERMAL SENSITIVITY

adaptation during range expansion because temporal variation in temperature differs latitudinally (see Chapter 2). Temporal variation should facilitate adaptation at low latitudes, but inhibit adaptation at high latitudes. Overall, migration appears to be a doubleedged sword. In small populations, immigrants help to prevent extinction during the process of adaptation. In moderate to large populations, however, the inﬂux of maladapted genotypes prevents perfect adaptation. Some degree of maladaptation will always persist when migration occurs. Thus, gene ﬂow initially enhances the probability of adaptation during a range expansion but reduces this probability once adaptation has begun.

3.10 Conclusions Throughout this chapter, we have seen that the current theory provides a limited understanding of the evolution of thermal sensitivity. Although many mechanisms of thermal adaptation should impose a specialist–generalist tradeoff, optimality models that incorporate this constraint often fail to explain empirical observations. The mismatch between theory and data seems particularly great for the thermal sensitivities of growth and development, which often follow patterns of countergradient or

87

cogradient variation. We cannot solely attribute the failures of the theory to a naïve view of thermal heterogeneity because reciprocal transplant and laboratory selection experiments have also revealed the failings of optimality models. We now have sufﬁcient evidence for the need to consider additional constraints when modeling ﬁtness landscapes, such as acquisition and allocation tradeoffs. We also need to pay more attention to evolutionary processes that limit rates of adaptation, such as mutation, drift, and migration. The slow rates of adaptation predicted by quantitative genetic and allelic models of performance curves might explain why experimental evolution at constant temperatures has not produced thermal specialists. Genetic models promise to guide our understanding of the limits to adaptation if we can obtain the data needed to parameterize these models for speciﬁc populations. Nevertheless, both optimality and genetic models of performance curves make critical assumptions that we know to be false. First, these models assume that organisms fail to thermoregulate in the face of severe thermal heterogeneity. Second, they assume the thermal sensitivity of performance cannot acclimate in response to temperatures experienced during ontogeny. A more advanced theory would need to account for these strategies of thermal adaptation. In route to such a theory, we next consider the evolution of thermoregulation and the evolution of thermal acclimation.

CHAPTER 4

Thermoregulation

Although environmental temperatures vary tremendously over space and time, most organisms regulate their body temperature through behavior, physiology, and morphology. By thermoregulation, I do not mean the temperature of an organism simply differs from that of its surroundings. Rather, a thermoregulator maintains a particular mean or variance of body temperature (Fig. 4.1), using neural mechanisms to sense and respond to its environment (Bicego et al. 2007). Although this phenomenon appears most pronounced in endotherms such as birds and mammals, virtually all organisms exhibit some degree of thermoregulation. For example, most organisms use behavior or physiology to avoid lethal extremes, and many ectothermic animals maintain a precise body temperature despite variation in environmental temperature (Beitinger and Fitzpatrick 1979; Brattstrom 1979; Danks 2004; Heinrich 1981; Huey 1982; Lagerspetz and Vainio 2006). Even slime molds will migrate toward a speciﬁc environmental temperature when provided with the opportunity (Whitaker and Poff 1980). In fact, the boundaries between poikilotherms and homeotherms have blurred considerably because we now know that many “homeotherms” exhibit diel and seasonal variation in body temperature (Prinzinger et al. 1991). In this chapter, we shall examine models of optimal thermoregulation and use these models to understand the behavior of organisms in natural and artiﬁcial environments. We begin by considering the potential beneﬁts and costs of thermoregulation and then focus on a model that maximizes the net energetic beneﬁts. We then expand our focus to include models that also consider nonenergetic beneﬁts and costs. Finally, we shall contemplate the origin and maintenance of endothermic thermoregulation. With minimal interpretation, models 88

of optimal thermoregulation apply equally well to ectotherms and endotherms.

4.1 Quantifying patterns of thermoregulation The earliest studies of thermoregulation were focused on the relationship between ambient temperature and body temperature. Using this approach, researchers attempted to distinguish between thermoregulators and thermoconformers (Fig. 4.2). A perfect thermoregulator would possess an invariant body temperature over a range of air or water temperatures. On the other hand, a perfect thermoconformer would possess a body temperature equivalent to the ambient temperature. Imperfect thermoregulators exhibit some intermediate pattern. Using these criteria, we would conclude that a variety of organisms thermoregulate to some degree under natural conditions, including insects (Harrison et al. 1996; Verdu et al. 2004), amphibians (Duellman and Trueb 1986), reptiles (Van Damme et al. 1990), and ﬁsh (Baird and Krueger 2003). More importantly, though, different species seem to adopt very different thermoregulatory strategies in similar environments, as evidenced by variation in the slope of this relationship among sympatric species (Fig. 4.3). Although the relationship between air temperature and body temperature has been the most widely used index of thermoregulation, Heath (1964) cautioned thermal biologists to carefully consider the meaning of this relationship. He placed beer cans (ﬁlled with water) in the sun and recorded their temperatures throughout the day. The cans were generally hotter than the surrounding air, despite their obvious lack of a thermoregulatory strategy (Fig. 4.4). Not surprisingly, the physical properties

T H E R M O R E G U L AT I O N

(a)

(a) Tb

Thermoregulator

Body temperature

Frequency

89

Te

r

me

or

f on

oc

rm he

T

(b)

Air temperature (b)

Te

Temperature Figure 4.1 (a) Thermoregulation can reduce the variance of body temperature (Tb ) relative to variance of operative environmental temperature (Te ). (b) Alternatively, thermoregulation can alter the mean and reduce the variance of body temperature relative to the mean and variance of operative environmental temperature.

Body temperature (°C)

Frequency

Tb

40 36

Inactive Active

32 28 24 20 16 16

18

20

22

24

26

28

30

32

34

Air temperature (°C)

of a can, or an organism, differ markedly from those of air. Heath’s message did not fall on deaf ears. By the 1970s, researchers were combining observations of behaviors and temperatures to deduce whether animals were actually thermoregulating (see, for example, Huey et al. 1977). And by the 1980s, many researchers began to examine thermoregulation more rigorously with the help of operative temperature models (see Chapter 2). In the simplest case, the temperatures of animals were compared with the operative temperatures of their surroundings (Christian et al. 1998; Van Damme et al. 1989). In more sophisticated efforts, researchers attempted to model the frequency distribution of operative temperatures (Grant and Dunham 1988, 1990; Newell and Quinn 2005). If the distribution of body temperatures deviated from the distribution of operative temperatures, researchers concluded that the organism thermoregulated in its natural environment (see Fig. 4.5). Still, we cannot know whether these patterns reﬂect a regulation of body temperature because other activities might better explain deviations between potential and realized

Figure 4.2 (a) The relationship between air temperature and organismal temperature might indicate some form of thermoregulation. (b) For example, the temperatures of active lizards fell within a narrow range, regardless of the air temperature. However, the temperatures of inactive lizards depended strongly on air temperature. Adapted from Bauwens et al. (1999) with permission from Blackwell Publishing.

body temperatures. For example, operative temperature could covary with another environmental resource, such as water, food, or refuge. For this reason, we need to eliminate confounding factors that occur in natural environments if we want to know something about the thermal preference of an organism. The best evidence of thermal preference comes from measuring the body temperatures of organisms in thermal arenas (Box 4.1). These arenas consist of linear gradients or joined compartments that offer an organism the ability to choose its temperature. Such simple, controlled environments offer the advantage that one can easily quantify the frequency

90

T H E R M A L A D A P TAT I O N

Body temperature (°C)

(a) 40

(b) 40

Aglais urticae

30

30

20

20

10

10

0

0 0

(c) 40

10

20

30

40

30

20

20

10

10

0

10

20

0

(d) 40

Polygonia c-album

30

0

Inachis io

30

0 40 0 Air temperature (°C)

10

20

30

40

30

40

Vanessa atalanta

10

20

Figure 4.3 Relationships between air temperature and body temperature varied greatly among species of butterﬂies and depended strongly on environmental conditions. On sunny days (open symbols), Aglais urticae and Inachis io maintained a narrow range of body temperature despite a broad range of air temperatures, whereas Polygonia c-album and Vanessa atalanta conformed to air temperatures. On cloudy days (ﬁlled symbols), temperatures of all four species conformed to air temperatures. Adapted from Bryant et al. (2000) with kind permission from Springer Science and Business Media. © Springer-Verlag 2000.

of operative temperatures. By placing individuals in an environment where only temperature varies over space, researchers have conﬁrmed distinct and repeatable thermal preferences. Although a variety of factors can inﬂuence the preferred body temperature, this measure can also be a remarkably repeatable property of a species. For example, Angilletta and Werner (1998) recorded a preferred body temperature for the gecko Christinus marmoratus that differed by less than a degree from the preferred body temperature recorded for the same species about three decades earlier (Licht et al. 1966b). Preferred body temperatures provide valuable information about thermoregulatory strategies in environments where no ecological constraints exist. By combining this information with measures of operative temperatures and body temperatures in natural environments, we can discover whether organisms regulate their temperature in circumstances where many factors can inﬂuence behavior.

Eventually, the development of thermal indices by Hertz and colleagues (1993) set a new standard for quantifying thermoregulation in natural environments (Box 4.2). The opportunity for thermoregulation was captured by the index de , which equals the difference between the mean operative temperature of the environment and an animal’s preferred temperature (or preferred range of temperatures). The accuracy of thermoregulation was captured by the index db , which equals the difference between the mean temperature of the animal and its preferred temperature. Variations on this theme have been proposed, but all preserve the idea that de adequately estimates the quality of the thermal environment. Although these thermoregulatory indices emerged relatively recently, thermal biologists have already documented considerable variation in the accuracy and effectiveness of thermoregulation in natural environments (see, for example, Christian and Weavers 1996).

T H E R M O R E G U L AT I O N

(a)

1 1 2 1 3 2

30

1 2 1 7

1 1 2 2 2 2 1

3 4 4

60

1

50

Temperature (°C)

1 3

2 1 1

3 1 3 1

3 2 3 2

1

2

3

1

3

Mean Te Mimium Te Maximum Te Snake Tb

40 30 20 10 0

1 1

2 1 4

(b)

2

0.10 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.00

30

35

Air temperature (°C) Figure 4.4 A difference in temperature between an organism and its environment does not necessarily imply thermoregulation. For example, the temperatures of beer cans differed from those of the surrounding air by as much as 8◦ C (numerals indicate the number of records at each temperature). This pattern likely resulted from the passive orientation of the cans to the changing angle of solar radiation; upright cans should have intercepted the most radiation during early and late hours of the day, when the air was relatively cool (Huey et al. 1977). From Heath (1964). Reprinted with permission from AAAS.

Accurate thermoregulation reﬂects the seemingly adaptive integration of behavioral, physiological, and morphological strategies. Yet, these strategies almost certainly involve costs that could offset the beneﬁts of thermoregulation. How can we determine the adaptive value of any degree of thermoregulation? To do so, we could take an optimality approach, as we did to study the evolution of thermal sensitivity (Chapter 3). Before we can develop an optimality model of thermoregulation, we must consider the kinds of beneﬁts and costs associated with thermoregulatory strategies.

4.2 Beneﬁts and costs of thermoregulation 4.2.1 Beneﬁts of thermoregulation Undoubtedly, thermoregulation evolved as a means of enhancing performance in variable environments. An organism that regulates its body temperature

2 0 –2 –4 –6

Observation difference Zero difference Probability

–8 –10 0

5

10 Hour

15

Probability

4

25

6 4

25

20

70

1

Observed difference

Can temperature (°C)

35

91

20

Figure 4.5 During August 1997, timber rattlesnakes in Arkansas maintained a distribution of body temperatures that differed from the distribution of operative temperatures. (a) The mean body temperature exceeded the mean operative temperature in the morning and the evening, but the reverse was true at midday. (b) The observed difference between the mean body temperature and the mean operative temperature (left axis, circular symbols) was statistically signiﬁcant at almost all hours of the day. This plot also shows the probability of observing an equal or greater difference by chance (right axis, square symbols). Given the number of comparisons, a probability of <0.002 corresponds to a type I error rate of 0.05. Adapted from Wills and Beaupre (2000, Physiological and Biochemical Zoology, University of Chicago Press). © 2000 by The University of Chicago.

can ensure a maximal (or nearly maximal) level of performance. Because thermoregulation buffers the heterogeneity of the environment, effective thermoregulators could also evolve greater performance through thermal specialization (see Chapter 3). Still, an important question remains. Given the potential to evolve a low thermal optimum for performance, why do many organisms maintain relatively high body temperatures? In other words, why regulate the mean and variance of body temperature when an organism could more easily regulate the variance and conform to the mean environmental temperature? The high body temperatures of endotherms and many ectotherms probably serve to overcome

T H E R M A L A D A P TAT I O N

Box 4.1 Measuring preferred body temperatures When placed in artiﬁcial thermal gradients, organisms generally maintain body temperatures within a narrow range. Because an organism in such an environment lacks physical and ecological constraints on thermoregulation, the temperature selected presumably reﬂects its preference, or the preferred body temperature. Although many authors consider the mean temperature as the preferred body temperature, a model based on upper and lower set points appears to better ﬁt the distributions of temperatures in these experiments (Barber and Crawford 1977). In this case, one might choose to consider a preferred range, rather than a single preferred temperature (see, for example, van Berkum et al. 1986). Preferred body temperatures must be measured in artiﬁcial conditions that enable an organism to select from a wide range of equally accessible microclimates. These conditions usually require a thermal arena, consisting of a linear (Licht et al. 1966b), circular (Bowker 1984), or patchy (Pulgar et al. 1999) distribution of operative temperatures. To create a uniform gradient, a conductive surface, such as a continuous metal ﬂoor, would be heated at one end and cooled at the other end. Aquatic gradients require more sophisticated designs than terrestrial ones (see Wallman and Bennett 2006). Generally, nonthermal resources (e.g., refuge) should occur uniformly within the gradient or should not occur at all; otherwise, movement within the gradient could reﬂect the compromised use of thermal and nonthermal resources. Controlling the distribution of nonthermal resources can be trickier than one might think. For example, lizards prefer a different temperature when exposed to a combination of visible and infrared radiation instead of infrared radiation alone (Sievert and Hutchison 1988). Likewise, snails prefer speciﬁc combinations of temperature and light intensity; in one experiment, snails thermoregulated well in steep gradients of visible light, but did not do so in shallow gradients of visible light (Rossetti and Cabanac 1985). Based on these and similar observations, we might choose to decouple these two forms of radiation within a thermal gradient; however, animals in natural environments generally bask in sunlight consisting of both visible and infrared wavelengths. Therefore, no gradient can simultaneously reﬂect the natural conditions of thermoregulation and eliminate the confounding effects of nonthermal resources. When in doubt, we could compare the movements of animals in an artiﬁcial thermal gradient with their movements in the same enclosure without the thermal gradient; no predictable pattern of movement should be observed when in the absence of the thermal

gradient (see, for example, Fig. 4.6). Alternatively, we could use a null model to predict the movements of a thermoconformer and compare this prediction with the observed behavior (Anderson et al. 2007). Once an organism enters a thermal gradient, it may spend some period habituating to artiﬁcial conditions before thermoregulating normally. Exploratory behavior could lead to inaccurate estimates of preferred body temperature unless recordings are postponed for some period. Often, researchers introduce individuals to the experimental conditions at least 1 day prior to measurements. If the period of habituation were uncertain, body temperature could be monitored continuously until ﬂuctuations appear to give way to a fairly stable value; this stable value after habituation has been termed the ﬁnal preferendum (Reynolds and Casterlin 1979). Once an individual habituates to the experimental conditions, it usually maintains a similar body temperature over the course of several days (Arnold et al. 1995; Dohm et al. 2001; Le Galliard et al. 2003), suggesting the preferred body temperature constitutes a real property of an organism. Furthermore, genetic variation in preferred body temperature has been documented through common garden experiments (Freidenburg and Skelly 2004) and laboratory natural selection (Good 1993).

16 Thermal gradient No gradient

14

Number of snails

92

12 10 8 6 4 2 0 14

16

18 20 22 Temperature (°C)

24

26

Figure 4.6 Thermoregulation can be inferred from the behaviors of animals in the presence or absence of a thermal gradient. For example, the distribution of snails in a thermal gradient became more uniform when the thermal gradient was turned off. Data from Gerald and Spezzano (2005).

T H E R M O R E G U L AT I O N

93

Box 4.2 Indices of thermoregulation

d E =1− b de

(4.1)

One can infer the relative intensity of thermoregulation from the value of E . Random use of thermal microclimates yields an E equal to zero and perfect regulation of body temperature yields an E equal to 1. To demonstrate the value of these indices, Hertz and colleagues (1993) compared the thermoregulatory behaviors of several species of Puerto Rican anoles. Figure 4.7 shows a subset of their data. In August, the

preferred temperatures of Anolis cristatellus occurred rather infrequently in its environment at high elevation (de = 5.0). Despite this fact, lizards maintained a mean body temperature that was closer to the set-point range than was the mean operative temperature (db = 2.5), resulting in a moderate effectiveness of thermoregulation (E = 0.5). In contrast, A. gundlachi experienced an environment in which its preferred temperatures occurred very frequently (de = 0.7). Consequently, the extremely accurate thermoregulation by this species (db = 0.6) could have easily resulted from random use of microenvironments, which is reﬂected in its low effectiveness of thermoregulation (E = 0.1). These data illustrate how thermoregulatory indices enable one to distinguish between a thermoregulator and a thermoconformer in a natural environment. Anolis gundlachi 210 m, August

Anolis cristatellus 1150 m, August 25

(a)

Tb

(b)

Tb

20 15 Frequency of observations

Hertz and colleagues (1993) developed a way to quantify thermoregulatory behavior, which combines information about the preference of body temperatures in an artiﬁcial thermal gradient and the availability of operative temperatures in a natural environment. They presented three questions relevant to the study of thermoregulation. First, how precisely does an organism regulate its body temperature when freed from ecological constraints? To answer this question, they used the range of body temperature selected in an artiﬁcial thermal gradient, called the set-point range. Most researchers consider the central 50% or 80% of preferred body temperatures to deﬁne the set-point range. Second, how accurately does an organism maintain its body temperature within the set-point range? To answer this question, they calculated an index known as the accuracy of thermoregulation (db ), which equals the mean absolute deviation of body temperature from the set-point range. The index db provides a way to compare the extent to which body temperatures in natural environments correspond to preferred temperatures. Finally, how well do thermoregulatory strategies actually enhance the accuracy of thermoregulation? This question arises because a low db could result from one of two causes: (1) an organism regulates it body temperature within its set-point range or (2) operative temperatures generally fall within the set-point range. To establish that accurate thermoregulation (i.e., a low db ) results from behavioral or physiological regulation, one must have some knowledge of the operative thermal environment. Armed with such knowledge, Hertz and colleagues calculated an index known as the thermal quality of the environment (de ), which equals the mean deviation of operative temperature from the set-point range. Using the values of db and de , they calculated a ﬁnal index known as the effectiveness of thermoregulation (E ):

10 5 0 25

(c)

Te

(d)

Te

20 15 10 5 0 15

20

25

30

35 40 15 20 Temperature (°C)

25

30

35 40

Figure 4.7 Thermoregulatory behavior differed between two species of lizards inhabiting different environments. (a–d) The distributions of body temperatures (Tb ) and operative temperatures (Te ) for Anolis cristatellus and A. gundlachi. The set-point range of each species is delineated by the shaded region. Anolis cristatellus thermoregulated both accurately and effectively, whereas A. gundlachi appeared to maintain its preferred temperature without thermoregulation. Adapted from Hertz et al. (1993, The American Naturalist, University of Chicago Press). © 1993 by The University of Chicago.

T H E R M A L A D A P TAT I O N

thermodynamic constraints on biochemical reactions. As discussed in Chapter 3, organisms in hot environments should evolve relatively stable enzymes. Certain biochemical structures provide functional stability at high temperatures, but they also slow the rate of catalysis, Kcat (Hochochka and Somero 2002). Nevertheless, the catalytic rate increases with increasing temperature. When the effects of enzymatic structure and environmental temperature combine, the advantage of a high body temperature more than outweighs the disadvantage of a stable enzymatic structure (Hochochka and Somero 2002; Somero 2004). In other words, hotter is better for the performance of enzymes.10 Scaling up this logic, an organism with a high thermal optimum should outperform an organism with a low thermal optimum. We can test this hypothesis using comparative or experimental data. Comparative studies of multiple clones or populations within species enable us to search for a positive relationship between the thermal optimum and the maximal performance. Selection experiments that cause the evolution of new thermal optima should also lead to correlated changes in maximal performance. With either kind of evidence, we can quantitatively test the idea that thermodynamics constrain organismal performance. Based on the current metabolic theory (Gillooly et al. 2001, 2002; Savage et al. 2004), the relationship between the thermal optimum (Topt ) and the maximal performance (Pmax ) should follow this expression: Pmax ∝ e−E/kTopt ,

(4.1)

where E equals the mean activation energy of rate-limiting reactions and k equals Boltzmann’s constant. How well do thermodynamic constraints on biochemical function translate into organismal performance? To my knowledge, Eppley (1972) 10

In the literature, the term “hotter is better” or “warmer is better” has been used in different contexts. For example, Bennett (1987b) used the term “warmer is better” to describe an increase in performance associated with an acute change in body temperature. In contrast, I use this term to describe an increase in maximal performance with an evolutionary change in the thermal optimum.

provided the ﬁrst evidence that hotter is better by comparing maximal growth rates of algal species (Fig. 4.8). In the decades that followed, many other comparative studies of performance curves accumulated. A survey of these relevant studies suggests that hotter is usually better (Table 4.1). Moreover, all four studies that compared rates of population growth (a common estimate of ﬁtness) strongly supported the existence of a thermodynamic constraint. Still, the magnitude of the thermodynamic effect often deviates signiﬁcantly from that predicted by the metabolic theory. For example, Frazier and colleagues (2006) compared rates of population growth for 65 species of insects. The effect of the thermal optimum on the maximal performance was greater than that predicted by the metabolic theory; in other words, hotter appeared even better than expected from thermodynamic considerations. If hotter is truly better, we should expect a correlated increase in maximal performance when

10

Detonula confervacea Chlorella pyrenoidosa Skeletonema costatum Dunaliella tertiolecta Ditylum brightwellii

8 Growth rate (doublings day–1)

94

6

4

2

0 0

10

20 30 Temperature (°C)

40

50

Figure 4.8 A comparative study of the thermal sensitivity of growth rate provided strong evidence of a thermodynamic effect on maximal performance. Data are population growth rates of ﬁves species of algae grown in laboratory cultures. Adapted from Eppley (1972) with permission from the National Marine Fisheries Service.

T H E R M O R E G U L AT I O N

95

Table 4.1 A review of comparative studies of thermal performance curves suggests that hotter is better. For each case, the correlation (r ) between the thermal optimum and the maximal performance is given, along with the number of populations or species used to estimate this correlation (N). Because correlations were computed on an Arrhenius scale, a negative correlation supports the hypothesis that hotter is better. Organisms

Performance

Frogs Marsh frogs Lizards Lizards Mites Fungal symbionts of lichen Algal symbionts of lichen Fish Arctic charr Trees Trees Pea aphids Endoparasites Wasps Algae Actinobacteria Water ﬂeas Insects

Locomotion2 Locomotion Locomotion2 Locomotion2 Locomotion3 Growth Growth Growth2 Growth2 Photosynthesis Photosynthesis Development Parasitization Parasitization Fitness Fitness Fitness Fitness

r

N

0.10 −0.96 −0.42 −0.21 0.39 0.53 −0.82 −0.10 −0.62 0.72 −0.18 −0.66 0.25 −0.31 −0.92 −0.85 −0.69 −0.504

51 5 101 7 5 4 7 23 11 8 8 5 12 26 4 6 29 65

Source Navas (1996a,b) Wilson (2001) Bauwens et al. (1995) van Berkum (1986) Deere and Chown (2006) Sun and Friedmann (2005) Sun and Friedmann (2005) Asbury and Angilletta (unpublished) Larsson et al. (2005) Gratani and Varone (2004) Cunningham and Read (2002) Lamb et al. (1987) Grewal et al. (1994) Carrière and Boivin (1997) Eppley (1972) Hahn and Pockl (2005) Palaima (2002) Frazier et al. (2006)

Notes: Fitness was estimated as the rate of population growth. 1 Some taxa were omitted because thermal optimum was not resolved. 2 Performance was adjusted for body size. 3 Performance was averaged among acclimation treatments. 4The correlation coefﬁcient was estimated from independent contrasts.

selection leads to a higher thermal optimum. Unfortunately, most selection experiments either did not include enough measures of performance to estimate thermal optima (Huey et al. 1991; James and Partridge 1995) or were too short to produce signiﬁcant shifts in thermal optima (Gilchrist et al. 1997). In an exceptional study, Santos (2007) allowed replicate populations of Drosophila subobscura to evolve at a constant temperature of 13◦ C, 18◦ C, or 22◦ C. He then measured the net ﬁtness of ﬂies from each line at all three temperatures. Not only did the thermal optimum diverge among these lines, but so did the mean ﬁtness. Consistent with the patterns observed in comparative studies, ﬂies that evolved at 22◦ C had the highest mean ﬁtness, whereas ﬂies that evolved at 13◦ C had the lowest. In an equally impressive study, Cooper and colleagues (2001) compared the thermal sensitivities of maximal growth rate between ancestral lines and selected lines of Escherichia coli after 20 000 generations at

37◦ C (see Chapter 3 for details). The ancestral lines grew maximally at 40◦ C. Not surprisingly, the selected lines experienced a decrease in their thermal optimum, such that their new optimum matched the temperature of their environment. Contrary to the hypothesis, the selected lines grew faster at their low thermal optimum than the ancestral lines did at their high thermal optimum. Furthermore, a study of mutational effects refuted a necessary link between the thermal optimum and the maximal ﬁtness; mutants of E. coli whose performance curves were shifted toward higher temperatures did not display a higher maximal ﬁtness relative to their ancestors (Mongold et al. 1999). Importantly, these patterns contrast the pattern derived by comparing two natural isolates of E. coli (Bronikowski et al. 2001). Hence, support for the hypothesis from comparative studies might be countered by evidence from selection experiments involving the same species.

96

T H E R M A L A D A P TAT I O N

Despite the mounting evidence that hotter is better, the thermodynamic effect differs considerably among species and performances (see Table 4.1). Furthermore, a comparative study of extreme thermophiles reafﬁrms that life has its limits; the maximal growth rate of cyanobacteria increased with an increasing thermal optimum up to 60◦ C, but decreased above this temperature (Miller and Castenholz 2000). This ﬁnding implies that extremely high temperatures impose a physical stress that outweighs any positive effect on catalytic rates. Clearly, we need more research to understand variation in the thermodynamic effect and the mechanisms that enable certain lineages to overcome this constraint. But for now, the majority of data suggest that organisms beneﬁt from adaptation to high temperatures combined with precise thermoregulation.

4.2.2 Costs of thermoregulation Although the beneﬁts of thermoregulation seem fairly straightforward, the costs can be quite complex. We shall consider three kinds of costs: energy expenditure, mortality risk, and missed opportunities. Both energy expenditure and mortality risk can easily be converted into demographic consequences. Energy expended for thermoregulation cannot be used to grow or reproduce. Thus, current and future fecundity would suffer during periods of intense thermoregulation. A greater risk of mortality would directly affect ﬁtness by reducing the likelihood of future reproduction. Missed opportunities seem more difﬁcult to quantify because one needs to know what activities conﬂict with thermoregulation and how these activities contribute to ﬁtness. If thermoregulation interferes with mating, an organism would suffer a decrement in fecundity. But a conﬂict between thermoregulation and feeding could reduce the survivorship or fecundity of thermoregulators. Cost–beneﬁt analyses of thermoregulation are complicated by the fact that organisms use many mechanisms to thermoregulate (Chappell and Whitman 1990; Lustick 1983), each of which imposes particular costs. Let us brieﬂy consider behavioral, physiological, and morphological mechanisms and the kinds of costs they can impose.

Behavioral mechanisms of thermoregulation are most readily discernable. Many kinds of animals shuttle between hot and cold patches to thermoregulate. For example, lizards preferentially perch on exposed rocks during the morning and evening but move under bushes during midday (Bauwens et al. 1996). Similarly, ﬁsh shuttle within thermally stratiﬁed waters (Snucins and Gunn 1995), and Japanese beetles shuttle between sun and shade (Kreuger and Potter 2001). Even plants can track the position of the sun to maximize their absorption of solar radiation (Ehleringer and Forseth 1980). When the body temperature of a species remains constant along a geographic cline, shifts in behavior clearly play an important role. The lizard Sceloporus occidentalis perches on trees at low elevation, its congener S. graciosus remains on the ground at high elevation, and both species prefer a mixture of microhabitats at mid-elevation (Adolph 1990). When transplanted to a common environment, lizards from all three elevations used the same microhabitats (Asbury and Adolph 2007). In contrast, genetic factors contribute to geographic differences in the thermoregulatory behaviors of some grasshoppers (Samietz et al. 2005). In addition to overt shuttling, small shifts in posture and orientation can facilitate radiative and convective heat exchange between an organism and its environment (Ayers and Shine 1997; Bauwens et al. 1996; Coelho 2001; Heath 1965; Heinrich 1990; Munoz et al. 2005). Behavioral thermoregulation imposes energy expenditure, mortality risk, and missed opportunities; for example, organisms shuttling between microclimates will spend more energy, attract more predators, and startle more prey than will organisms remaining still (Huey 1974, 1982; Huey and Slatkin 1976). Physiological mechanisms—including circulation, evaporation, and metabolism—offer more subtle means of thermoregulation (Bartholomew 1982; Kammer 1981). The ﬂow of blood (or hemolymph) distributes heat between the core and surface of the body, determining rates of conduction and convection (Grigg and Seebacher 1999; Seebacher and Grigg 2001; Smith 1979). Organisms can quickly lose heat by shunting blood to body parts with high surface area relative to volume, such as the head or limbs (Dzialowski and O’Connor 2004; May 1995).

T H E R M O R E G U L AT I O N

(Bartholomew 1981; Stevenson 1985). For example, the use of blood ﬂow to regulate heating and cooling requires a substantial size (see Dzialowski and O’Connor 2001). Furthermore, larger insects generate more metabolic heat during ﬂight (Bishop and Armbruster 1999). Very large organisms possess a thermal inertia that dampens ﬂuctuations in body temperature and could even result in homeothermy without effort (Fig. 4.9, see also Seebacher et al. 1999). But bigger is not always better; small butterﬂies can use postural changes to alter radiative and convective heat ﬂuxes more effectively than large butterﬂies (Kemp and Krockenberger 2004). Morphologies associated with thermoregulation, such as color and size, require energy to produce and maintain. Dark color results from the deposition of pigments in the skin, fur, or feathers (Cooper and Greenberg 1992; Riley 1997). Organisms must incur some cost of producing pigments for thermoregulation, given the loss of melanin associated with poor diet (Talloen et al. 2004) and life underground (Jeffery 2006; Mejia-Ortiz and Hartnoll 2005; Parzefall 2001). Large size requires energy for growth and maintenance at the expense of other activities. Even when a small size facilitates thermoregulation, organisms can save energy on growth but might suffer loss of performance in foraging,

40 0.01 Temperature (°C)

Greater cardiac output demands more energy, and shunting of blood likely alters the metabolic capacity of peripheral tissues. Alternatively, countercurrent systems enable the ﬂow of blood to the periphery while retaining heat within the core (Holland et al. 1992). Through evaporation, organisms can dump heat at the expense of water. Because water loss affects performance (reviewed by Chown et al. 1999; Lillywhite and Navas 2006; Tsai et al. 1998; Walvoord 2003), organisms tend to reserve evaporation as a strategy of last resort; for instance, waterproof frogs tracked ambient temperature up to 40◦ C and then used evaporative water loss to prevent further increases (Shoemaker et al. 1987). To aid in evaporative cooling, organisms can seek out regions of low humidity when exposed to extreme heat (Prange and Hamilton 1992). Behaviors, such as panting or tongue lashing, can also enhance the rate of evaporation (Roberts and Harrison 1998). Finally, metabolism generates heat at the expense of chemical energy. This strategy enables endothermy in plants (Patino et al. 2000), insects (Heinrich 1974), ﬁsh (Block and Finnerty 1994), birds (Prinzinger et al. 1991), and mammals (Grigg et al. 2004). Bees increase the frequency of their wingbeats as the temperature of the environment decreases (Harrison et al. 1996; Roberts and Harrison 1998). Likewise, birds and mammals undergo shivering thermogenesis when their body temperature drops below a critical level (Scholander et al. 1950a). The costs of physiological thermoregulation depend on the speciﬁc mechanism, but include both energy expenditure and mortality risk. Morphology also serves to regulate body temperature. Changes in color—whether rapid, seasonal, or developmental—enable organisms to adjust their solar absorptivity (Clusella-Trullas et al. 2008; Hazel 2002; Kingsolver and Huey 1998; Lacey and Herr 2005; Walton and Bennett 1993). Adjusting the position of fur or feathers alters heat exchange by convection and radiation (Cossins and Bowler 1987; Scholander et al. 1950b). Even when morphology changes slowly, the current morphology plays an indirect role in thermoregulation by inﬂuencing the effectiveness of behavioral and physiological strategies. Size and shape have the greatest impact on the capacity for thermoregulation

97

0.1 35

0.5

1

30

2.5 5 kg

25 20 0

20

40

60

80

Time (min) Figure 4.9 The rates of heating and cooling depend on the mass of an organism. This plot shows calculated rates of heating for a range of body masses. Large organisms (≥5 kg) heat and cool so slowly they experience inertial homeothermy. Adapted from Seebacher and Shine (2004, Physiological and Biochemical Zoology, University of Chicago Press). © 2004 by The University of Chicago.

98

T H E R M A L A D A P TAT I O N

defense, or reproduction. Therefore, morphological traits designed for thermoregulation can impose energy expenditure, mortality risk, and missed opportunities. Importantly, organisms generally thermoregulate by a combination of behavior, physiology, and morphology, switching rapidly between suites of strategies (Box 4.3). In fact, the same pattern of thermoregulation can arise from different mechanisms in closely related organisms. Crepuscular cicadas thermoregulate using metabolic heat, but diurnal cicadas thermoregulate using behavior (Sanborn et al. 2004). This complexity challenges our abilities to understand the evolution of thermoregulation. Although the gross beneﬁts of thermoregulation depend on body temperature alone, the costs of thermoregulation depend on the suite of mechanisms that an organism adopts in a particular environment.

4.3 An optimality model of thermoregulation The prevailing theory of thermoregulation stems from Huey and Slatkin (1976), who modeled the optimal degree of thermoregulation given energetic costs and beneﬁts. They reasoned the cost of thermoregulation could outweigh the beneﬁts, particularly in environments where thermal resources were scarce. This argument was formalized in an optimality model, which predicts thermoregulatory behavior when the energetic costs and beneﬁts are known. Additional costs, such as predation risk and missed opportunities, were also discussed but were not modeled explicitly (Huey and Slatkin 1976; Huey 1982). Here, I shall derive their original model focusing on energetic costs and beneﬁts. The model captures the thermoregulatory strategy of an organism with a parameter k, which

Box 4.3 Thermoregulation by butterﬂies: a case study of integrated mechanisms Lepidopterans rely on a combination of behavior, physiology, and morphology to thermoregulate. As larvae, individuals develop their color based on the environmental temperature (Nice and Fordyce 2006). Dark coloration, provided by melanin, helps caterpillars in cold environments to raise their temperature through basking. As their body temperature rises, they seek shade within vegetation and also engage in physiological mechanisms of cooling (Fig. 4.10). As adults, butterﬂies combine wing melanism with postural adjustments to raise their temperature prior to ﬂight. Three forms of basking have been described: dorsal, lateral, and reﬂectance basking (Casey 1981; Kemp and Krockenberger 2002). Each of these strategies relies on an interplay between behavior and morphology for effectiveness (Kingsolver 1987). During dorsal basking, an individual spreads it wings ﬂat to allow solar radiation to fall directly on the thorax. Most of the heat gain occurs as the individual absorbs solar radiation at the thorax or conducts heat to the thorax from the basal portions of the wings. Therefore, melanism on the basal regions of the dorsal surface will speed heating during dorsal basking. During lateral basking, an individual closes its wings and orients perpendicular to solar radiation. In this position, melanin in the basal regions of the ventral surfaces enhances heating. During reﬂectance basking, an individual opens its wings at an angle to reﬂect solar radiation from

the dorsal surface onto the thorax. A lack of melanin along distal regions of the wings enhances reﬂection. Heinrich (1990) disputed the signiﬁcance of reﬂective basking, suggesting that heating was instead mediated by dorsal basking, with elevated wings to minimize convection. Regardless, this alternative interpretation still involves the integration of behavior and morphology. Faster rate of blood flow Crawling Convective cooling Descent to the base of the tree in search of cooler microhabitat

Sun seeking 20

22

Avoidance of sun 24

26 28 30 32 Air temperature (°C)

34

36

Figure 4.10 An ethogram depicts the suite of mechanisms used for thermoregulation. This ethogram was constructed for larvae of the moth Antheraea assama. Arrows depict transitions between thermoregulatory mechanisms. Reproduced from Bardoloi and Hazarika (1994) with permission from Blackwell Publishing.

T H E R M O R E G U L AT I O N

density function for operative temperature, f (Te ); this function describes the chances of encountering speciﬁc operative temperatures. Integrating the beneﬁt of thermoregulation weighted by the probability density function yields the expected beneﬁt of ¯ thermoregulation (b): b¯ = b(Tb (k, Te )) · f (Te )dTe . (4.4)

determines how closely the body temperature (Tb ) tracks the environmental temperature. Originally, Huey and Slatkin considered ambient temperature, but we can replace this variable with the operative temperature (Te ). Therefore, k determines thermoregulation according to the following relationship: Tb = Topt + k(Te − Topt ).

(4.2)

An organism that maintains a body temperature equal to its thermal optimum (i.e., k = 0) would maximize its expected beneﬁt. However, we should not necessarily expect perfect thermoregulation to evolve because this strategy doubtless involves costs. The optimal strategy would maximize the net beneﬁt, which equals the expected beneﬁt minus the expected cost. Huey and Slatkin modeled the cost of thermoregulation (c) as a function of the difference between the target temperature of the organism and the operative temperature of the microenvironment (Fig. 4.11c):

By choosing a value of k between 0 and 1, we can generate strategies that range from perfect thermoregulation to perfect thermoconformation (Fig. 4.11a). The beneﬁt of thermoregulation (b) depends on the body temperature of the organism (Fig. 4.11b). Because the body temperature depends on the thermoregulatory strategy and operative temperature, we can deﬁne the beneﬁt of thermoregulation more explicitly as b(Tb (k, Te )).

(4.3)

k = 0.6 k=1 Topt

(4.5)

Because the target body temperature depends on the thermoregulatory strategy and the operative temperature, we can deﬁne the cost of thermoregulation

(b)

Hypothetical benefit curves

(c)

Specialist

Energy/time

k=0

c(Tb − Te ).

Energy/time

Thermoregulatory strategies

Body temperature (Tb)

This beneﬁt represents the energy gained from adopting strategy k in the microenvironment characterized by a certain operative temperature. Because temperature varies spatially, we must use information about the thermal macroenvironment to calculate the expected beneﬁt of thermoregulation. The expected beneﬁt depends on the probability

(a)

99

Generalist

Hypothetical cost curves

High cost

Low cost Operative environmental temperature (Te)

Tb

Tb – Te

Figure 4.11 Modeling the optimal strategy of thermoregulation requires precise deﬁnitions of the potential strategies and their respective costs and beneﬁts. (a) The thermoregulatory strategy was modeled as a linear relationship between the operative environmental temperature (Te ) and the body temperature (Tb ). The slope of this relationship (k ) varied between 0 and 1, with a slope of 0 deﬁning a perfect thermoregulator and a slope of 1 deﬁning a perfect thermoconformer. (b) The rate of energy gain was assumed to increase as the body temperature approached the thermal optimum; two hypothetical curves are shown. (c) The energetic cost of thermoregulation was assumed to increase as the operative temperature deviated from the target body temperature; two hypothetical curves are shown. Adapted from Huey and Slatkin (1976, The Quarterly Review of Biology, University of Chicago Press). © 1976 by The University of Chicago.

T H E R M A L A D A P TAT I O N

c(Tb (k, Te ) − Te ).

(4.6)

This cost represents the energy required to reach a microenvironment in which Te = Tb (k, Te ), given the operative temperature of the current microenvironment. As with the expected beneﬁt, the expected cost of thermoregulation (¯c) for a particular macroenvironment must be calculated by integrating the weighted cost over all possible microenvironments: c¯ = c(Tb (k, Te ) − Te ) · f (Te )dTe . (4.7) Thus, the expected cost depends on the thermoregulatory strategy of the organism and the probability densities of operative temperatures. The thermoregulatory strategy determines the target body temperature that an organism will try to attain when faced with a given operative temperature. The probability density function determines the chance of encountering particular operative temperatures. Additionally, the energetic cost of thermoregulation depends on other properties of the organism and its environment. With respect to the organism, locomotor ability, thermal perception, and spatial memory will affect the energetic cost of locating a preferred microclimate. With respect to the environment, the size, connectivity, and autocorrelation of thermal patches will also matter (Fig. 4.12). Once we have formulated the expected beneﬁt and cost of a thermoregulatory strategy, we can calculate the net beneﬁt (B): b(Tb (k, Te )) − c(Tb (k, Te ) − Te ) B = b¯ − c¯ = · f (Te )dTe .

(4.8)

All else being equal, the optimal thermoregulatory strategy maximizes the net beneﬁt. By varying certain parameters and holding all others constant, Huey and Slatkin explored the optimal strategy k for various forms of b and c. Take another look at the functions shown in Fig. 4.11. By superimposing those functions for b and c on a single plot, the net beneﬁt of a perfect thermoregulator or a perfect thermoconformer can be depicted graphically (Fig. 4.13). From such plots, we can assess whether perfect thermoregulation, perfect conformation, or

Cost (energy/time)

(a)

more explicitly as:

Positive spatial autocorrelation

(b)

Negative spatial autocorrelation

Tb – Te

Tb – Te

Figure 4.12 The cost curve associated with thermoregulation depends on the spatial structure of the thermal environment. (a) With positive spatial autocorrelation, the current microenvironment of the organism likely resembles the surrounding microenvironments. Consequently, an organism in an undesirable microenvironment (large value of Tb − Te ) would need to move a great distance to locate a preferred microclimate. (b) With negative spatial autocorrelation, the current microenvironment of the organism likely differs greatly from the surrounding microenvironments. Consequently, an organism in an undesirable microenvironment would need to move only a short distance to locate a preferred microenvironment.

b(Topt) Energy gain (J h–1)

100

0

Benefit (a)

(b) Benefit

Cost

Environmental temperature (Te)

Topt

Environmental temperature (Te)

Topt

Figure 4.13 The beneﬁt and cost of thermoregulation depend on the strategy adopted by the organism. These hypothetical curves are based on the functions shown in Fig. 4.11. The hatched area depicts the difference between the beneﬁt and the cost. (a) A perfect thermoregulator (k = 0) will experience the same beneﬁt, b(Topt ), in all microenvironments, but will experience a greater cost as the operative temperature of the microenvironment (Te ) falls below the thermal optimum for energy gain (Topt ). (b) A perfect thermoconformer (k = 1) will experience no cost in any microenvironment, but will experience less beneﬁt as the operative temperature of the microenvironment falls below the thermal optimum for energy gain. Adapted from Huey and Slatkin (1976, The Quarterly Review of Biology, University of Chicago Press). © 1976 by The University of Chicago.

T H E R M O R E G U L AT I O N

(d) 1

2

Frequency

Frequency

(a)

Environmental temperature (Te) (b)

(e)

2

b1 = b2

Benefit

Benefit

b1(0,Te)

1

Environmental temperature (Te)

b2(0,Te)

Body temperature (Tb)

Body temperature (Tb) (f)

Cost

(c)

Cost

some intermediate strategy would maximize the net rate of energy gain. Using this model, Huey and Slatkin reasoned a decrease in the environmental temperature should cause a decrease in the mean body temperature of an optimal thermoregulator. This prediction follows from their choice of phenotypic function: recall, a simple linear function was used to describe the relationship between environmental temperature and body temperature (eqn 4.2). Consequently, mean body temperature must decrease as mean environmental temperature decreases, unless both warm and cold environments favor perfect thermoregulation. Broad comparative surveys of amphibians and reptiles support this prediction (Andrews 1998; Feder and Lynch 1982). Nevertheless, we should infer the cause of these patterns with caution. If the beneﬁt and cost curves vary among environments, optimal thermoregulation within environments would not necessarily lead to the anticipated pattern. Moreover, we do not know whether individuals in cold environments exhibit low body temperatures because of optimal thermoregulation or because of thermal constraints. For example, consider two environments that differ in their mean operative temperatures: a warmer environment that enables an organism to reach its thermal optimum for performance, and a colder environment that does not (Fig. 4.14a). Although an individual could thermoregulate perfectly (k = 0) in either environment, this strategy would confer a greater gross beneﬁt in the warm environment than it would in the cold environment (Fig. 4.14b). Clearly, the gross beneﬁt of thermoregulation depends on the range of body temperatures that an organism can attain in its current environment. The same conclusion holds for the cost of thermoregulation. When the thermal optimum cannot be attained, the cost curve for a perfect thermoregulator would be referenced to the closest attainable body temperature (Fig. 4.14c). Even if both environments did enable an individual to attain its thermal optimum, the cost of thermoregulation would still differ between environments. The environment in which the thermal optimum occurs more frequently should impose a lesser cost of thermoregulation (Fig. 4.14d–f). Huey and Slatkin did not envision this situation because they assumed the cost curve did not depend on the probability

101

1

2

2

1 Tb – Te

Tb – Te

Figure 4.14 The realized beneﬁt and cost of thermoregulation depend on the distribution of operative temperatures (Te ) in the environment. (a) Distributions of operative temperatures are shown for two hypothetical environments. The colder environment (1) does not permit the organism to attain its thermal optimum. (b) Because of constraints on body temperature, the realized beneﬁt of perfect thermoregulation in the warm environment, b2 (0, Te ), exceeds the realized beneﬁt in the cold environment, b1 (0, Te ). (c) The cost of thermoregulation in the cold environment (1) exceeds the cost in the warm environment (2). Note that the cost curve for the cold environment is shifted to reﬂect the difference in the target body temperature during thermoregulation. (d–f) These plots are the same as (a–c), except now both environments permit the organism to attain its thermal optimum, and thus b1 (0, Te ) = b2 (0, Te ). In (f), the difference between the two cost curves reﬂects the higher frequency of preferred microclimates in the warm environment (2), relative to the cold environment (1).

distribution of operative temperatures (see eqn 4.5). In general, beneﬁt and cost curves should vary considerably along latitudinal and altitudinal clines. To make useful predictions from the model, we must know the precise forms of these curves. Better descriptions of beneﬁts and costs are also needed for experimental tests of the model. For example, Herczeg and colleagues (2006) applied

102

T H E R M A L A D A P TAT I O N

Huey and Slatkin’s model to explain variation in thermoregulation by lizards in three different experimental arenas (Fig. 4.15). In arenas where most of the operative temperatures were either well above or well below the preferred range, lizards thermoregulated very accurately and effectively (mean db ≤ 1 and mean E > 0.8). Given the high values of de , lizards presumably thermoregulated by shuttling between hot and cold patches. In an arena where all of the operative temperatures were below the preferred range, lizards thermoregulated least accurately and effectively (mean db = 4 and mean E = 0.05); because lizards in the cold arena could not attain a preferred temperature no matter how they behaved, Herczeg and colleagues wisely calculated db and E using the closest available temperature to the set-point range. Given these results, we might conclude lizards in the coldest arena chose not to thermoregulate effectively because the cost of thermoregulation was too high. But this interpretation ignores the fact that the cost and beneﬁt of thermoregulation depend on thermal constraints. Assuming the thermal optimum for performance lies within the set-point range, the gross beneﬁt of thermoregulation in the coldest arena was lower than the gross beneﬁts in the other two arenas (see

= Accuracy of thermoregulation = Thermal quality of habitat

Absolute deviation (°C)

12 10 8 6 4 2 0 Hot

Warm

Cold

Figure 4.15 Indices of thermoregulation for Zootoca vivipara in three different thermal arenas: hot, warm, and cold. In the cold arena, indices were calculated by using the maximal operative temperature instead of the lower boundary of the set-point range. Data are medians ± quartiles of the accuracy of thermoregulation (db ) and the thermal quality of habitat (de ). Adapted from Herczeg et al. (2006) with kind permission from Springer Science and Business Media. © Springer-Verlag 2006.

Fig. 4.14b). Furthermore, the costs likely differed among arenas because a lizard in the cold arena could not have attained its thermal optimum (see Fig. 4.14c). Without better estimates of the beneﬁt and cost curves for lizards in these speciﬁc arenas, we cannot fully understand the behaviors documented in this experiment. In summary, thermal optima and operative temperatures alone cannot tell us the optimal strategy of thermoregulation because we need additional information to calculate the beneﬁts and costs of thermoregulation. Unfortunately, the paucity of additional information limits our ability to apply this optimality model to the real world. Despite this problem, we shall do our best to use Huey and Slatkin’s model to understand the thermoregulatory behaviors of organisms in artiﬁcial and natural environments.

4.4 Do organisms thermoregulate more precisely when the beneﬁts are greater? The optimal level of thermoregulation increases (k → 0) as the gross beneﬁt of thermoregulation (b) increases. Two properties of a performance curve determine the relative beneﬁts of thermoregulation and thermoconformation: the performance breadth and the maximal performance (see Fig. 3.4). As the performance breadth increases, small deviations of body temperature from the thermal optimum impose a lesser loss of energy gain. For a given cost curve, the net beneﬁt of thermoconformation for a generalist would exceed that for a specialist when the environmental temperature differs from the thermal optimum; consequently, a generalist should thermoregulate less precisely than a specialist. Furthermore, a decrease in the maximal performance (energy gain) decreases the net beneﬁt of thermoregulation at all environmental temperatures, which also tends to favor less precise thermoregulation. Therefore, a change in the shape or height of the performance curve can shift the optimal strategy away from perfect thermoregulation. Is there any empirical evidence for these phenomena? Despite the hundreds (if not thousands) of performance curves that have been quantiﬁed, we currently lack the data to know whether generalists

T H E R M O R E G U L AT I O N

and specialists thermoregulate differently. In cases where we know something about thermoregulatory strategies, we generally do not know the thermal sensitivities of energy gain. In cases where we know the thermal sensitivities, we generally do not know enough about the thermoregulatory strategies. We cannot merely compare the body temperatures of specialists and generalists because they usually inhabit different environments. Consequently, we must account for the different costs and constraints as well as the beneﬁts. Existing data enable us to explore the effect of maximal performance on thermoregulatory behavior with greater conﬁdence. The rate of energy gain at any temperature depends on the quantity and quality of food, which depend on the interactions among the thermoregulator, its competitors, and its prey. Although exact rates of energy gain are difﬁcult to measure under natural conditions, maximal rates can easily be manipulated in artiﬁcial or natural environments. In the simplest way, we could measure the preferred body temperatures of animals that were systematically fed or fasted (Regal 1966). Because starved animals gain no energy from thermoregulation, we should expect fasting to alter the behavior of an animal in a thermal gradient. Numerous investigators have compared the preferred body temperatures of recently fed individuals with those of fasted individuals (Table 4.2). In general, fasted individuals select lower body temperatures than do fed individuals. In one experiment in which several levels of ration were used (Mac 1985), the mean body temperature in a thermal gradient decreased as the ration decreased; this graded response accords with Huey and Slatkin’s model. One of two factors could explain these effects of fasting on thermoregulation (Huey 1982). First, fasted individuals gain less energetic beneﬁt from thermoregulation. Alternatively (and perhaps more likely), the optimal temperature for energy gain decreases as the rate of feeding decreases (see Fig. 3.3). In either case, these phenomena suggest organisms respond rapidly to changes in their thermal sensitivity of performance. The thermal responses to feeding observed in artiﬁcial thermal gradients should lead us to expect similar behaviors under natural conditions. Two teams of researchers have supplementally fed animals and recorded their body temperatures in natural

103

environments. Both of these experiments focused on snakes, whose body temperatures were recorded intermittently by radiotelemetry. Because these experiments took place in natural environments, the researchers also had to consider thermal heterogeneity over time; to do so, they compared both body temperatures and indices of thermoregulatory effectiveness (see Box 4.2) before and after feeding. After Brown and Weatherhead (2000) fed water snakes (Nerodia sipedon), the snakes did not increase their body temperature or their effectiveness of thermoregulation (E). These same researchers also observed snakes in an artiﬁcial thermal gradient, and saw no signiﬁcant change in body temperature associated with feeding. Both results contrast reports by other researchers who did observe increases in body temperature ranging from 2◦ C to 8◦ C after feeding (Lutterschmidt and Reinert 1990; Sievert and Andreadis 1999). Likewise, after BlouinDemers and Weatherhead (2001) fed rat snakes (Pantherophis obsoleta), the snakes did not maintain higher body temperatures relative to operative temperatures. This result contrasts the increase in preferred body temperature after feeding that was observed by the same researchers (see Table 4.1). Nevertheless, snakes in their natural environment did increase their effectiveness of thermoregulation after feeding, suggesting they had worked harder to maintain the same body temperature. Certain snakes shifted habitats after feeding, apparently to facilitate thermoregulation. Of 14 snakes that were fed in the interior of the forest, all but three shifted to the edge of the forest. In contrast, only two snakes that were fed at the edge moved to the interior. The differences between observations in artiﬁcial and natural environments could reﬂect uncontrolled variation in feeding rate or thermoregulatory cost, which cannot be avoided in natural environments. The quality of food also alters the beneﬁt of thermoregulation. By deﬁnition, food of low quality results in slower energy gain than food of high quality. Thus, we should expect animals eating food of low quality to thermoregulate less precisely than animals eating food of high quality. This hypothesis has been tested experimentally in both artiﬁcial and natural environments. Nussear and colleagues (1998) fed chuckwallas either a high-ﬁber diet or a low-ﬁber diet and measured their body

104

T H E R M A L A D A P TAT I O N

Table 4.2 Starvation generally lowers the preferred body temperatures of ectotherms. Species Insects Panstrongylus megistus Triatoma infestans Fishes Dasyatis sabina Girella laevifrons Rutilus rutilus Salmo salar Salvelinus namaycush Amphibians Bufo marinus Bufo woodhousii Desmognathus fuscus Triturus dobrogicus Reptiles Anolis carolinensis Charina bottae Morelia spilota Nerodia sipedon Nerodia sipedon Nerodia sipedon Pantherophis guttatus Pantherophis obsoleta Thamnophis sirtalis Trimeresurus stejnegeri

Effect of fasting on preferred temperature

Source

Decreased ≈0.1◦ C per day Decreased by ≈0.25◦ C per day

Pires et al. (2002) Minoli and Lazzari (2003)

1◦ C lower Selected higher temperatures more frequently and lower temperatures less frequently No effect on temperature during photophase; temperature during scotophase decreased by ≈0.2◦ C per day 2◦ C higher Decreased linearly with decreasing ration

Wallman and Bennett (2006) Pulgar et al. (1999)

No effect No effect during most times of day, but 2◦ C lower during early scotophase (1900–2300 h) No effect 2–3◦ C lower

Mullens and Hutchison (1992) Witters and Sievert (2001)

0.6◦ C lower in females and 1.3◦ C lower in males 2◦ C lower 3◦ C lower 8◦ C lower 2◦ C lower No effect 6◦ C lower 2◦ C lower 4◦ C lower in one subspecies, but no effect in another subspecies 5◦ C lower when in a gradient with water and refugia; no effect in an empty gradient

Brown and Grifﬁn (2005) Dorcas et al. (1997) Slip and Shine (1988) Lutterschmidt and Reinert (1990) Sievert and Andreadis (1999) Brown and Weatherhead (2000) Sievert et al. (2005) Blouin-Demers and Weatherhead (2001) Lysenko and Gillis (1980)

temperatures in a thermal gradient. In contrast to our expectation, the preferred body temperature did not differ between lizards on different diets. In a ﬁeld experiment, Underwood (1991) recorded the thoracic temperatures of honeybees as they arrived at feeders containing concentrated syrup or dilute syrup. As expected, bees were hotter during ﬂight when they fed on the more concentrated syrup. Finally, Pulgar and colleagues (2003) fed ﬁsh either a high-quality diet of bivalves or a low-quality diet of algae and recorded their preferred body temperatures. Fish consuming bivalves selected higher temperatures (16–19◦ C) than did ﬁsh consuming algae (10–13◦ C). Importantly, ﬁsh not subjected to a thermal gradient did not prefer any particular

van Dijk et al. (2002) Morgan and Metcalfe (2001) Mac (1985)

Moore and Sievert (2001) Gvoždík (2003)

Tsai and Tu (2005)

portion of the aquarium, suggesting the ﬁsh were actually regulating their temperature during the experiment. Energy gain relies not just on the availability of food, but also on the ATP required to fuel its capture, ingestion, digestion, and absorption. Under conditions of hypoxia, the aerobic capacity of an organism could fall below that required to sustain metabolic activities. All else being equal, a shortage of oxygen would reduce the beneﬁt of thermoregulation, as does a shortage of food. Based on this argument, an organism should select a lower body temperature in a hypoxic environment to minimize metabolic demands. Experimental studies of preferred body temperature strongly support this

T H E R M O R E G U L AT I O N

120 Relative performance (%)

prediction. Hypoxia reduces the preferred body temperatures of crustaceans, ﬁsh, and reptiles (Petersen et al. 2003; Rausch et al. 2000; Schurmann et al. 1991; Wiggins and Frappell 2000). Furthermore, the effect of hypoxia on thermoregulation depends on the levels of hemoglobin present in tissues; water ﬂeas with abundant hemoglobin exhibited a small reduction in preferred body temperature during hypoxia compared to water ﬂeas with less hemoglobin (Wiggins and Frappell 2000). Taken together, manipulations of food availability, diet quality, and oxygen concentrations conﬁrmed that ectotherms do respond rapidly to changes in the energetic beneﬁts of thermoregulation. Additional tests of this hypothesis could involve indirect manipulations of energy gain by changing the number of competitors. Although we have only meager experimental evidence of the indirect effects of competition on thermoregulation (see, for example, Diego-Rasilla and Perez-Mellado 2000), these effects should parallel those caused by direct manipulation of food intake (Huey and Slatkin 1976). Detecting the indirect effects of competition will require adequate control for temporal variation in environmental temperatures and directs effects of intraspeciﬁc aggression (see Sections 4.6 and 4.7)

105

100 80 60 40 20 0 10

14

18 22 26 30 34 Body temperature (°C)

38

42

Figure 4.16 Thermal sensitivities of sprint speed (solid line), endurance (dotted line), and metabolizable energy intake (dashed line) in the eastern fence lizard, Sceloporus undulatus. The distribution of preferred body temperatures in this species is depicted by the solid bars (scale is relative). Adapted from Angilletta et al. (2002b) with permission from Elsevier. © 2002 Elsevier Ltd.

every performance within a narrow range of temperatures (Huey 1982). How should organisms behave when the thermal optimum or performance breadth differs between two or more performances? To answer this question, we must consider the conﬂicts and compromises that arise when maximizing ﬁtness. Two examples will shed light on this topic: thermoregulation during infection and thermoregulation during pregnancy.

4.5 Nonenergetic beneﬁts of thermoregulation Although energy gain clearly depends on body temperature, thermoregulation offers beneﬁts that cannot be quantiﬁed in terms of energy. Just consider the variety of performances that depend on body temperature (see Table 3.1). Nonenergetic performances (e.g., development, locomotion, and immunocompetence) mediate indirect effects of body temperature on survivorship or fecundity. Depending on the environmental context, any or all of these performances could be considered beneﬁts of thermoregulation. For example, male cicadas use endothermy to enhance mate calling at night, when atmospheric conditions promote acoustic communication (Sanborn et al. 2003). When a single temperature maximizes both energetic and nonenergetic beneﬁts (Fig. 4.16), a simple cost–beneﬁt model will sufﬁce. But organisms cannot always maximize

4.5.1 Thermoregulation during infection Infection often elicits a change in the thermoregulatory strategy of an animal. In many cases, higher body temperatures are preferred—a phenomenon generally referred to as fever. Both endotherms and ectotherms are capable of sustaining fever, the latter group using behavioral rather than physiological mechanisms (see, for example, Sherman et al. 1991). Substantial evidence supports the view that fever enhances a host’s immunocompetence and impairs a pathogen’s growth (reviewed by Kluger 1979; Kluger et al. 1998). For example, Ouedraogo and colleagues (2004) noted that locusts (Locusta migratoria) infected by a fungus basked more frequently than did uninfected locusts. By manipulating the potential for basking, these researchers quantiﬁed a substantial increase in survivorship

106

T H E R M A L A D A P TAT I O N

associated with thermoregulation. The longer an infected locust could thermoregulate, the more likely it was to survive the course of the experiment (21 days). Further studies provided evidence that thermoregulation enhanced survivorship by raising hemocyte levels and impairing fungal growth (Ouedraogo et al. 2003). Given the phylogenetically widespread occurrence of fever (see Kluger 1979), the thermal optimum for immunological performance probably exceeds that for other performances; if so, fever could represent a shift in the optimal thermoregulation caused by a gain in survivorship that more than compensates for the loss of other performances. Likely, a more complex relationship exists between temperature and host–pathogen interactions. Although high temperatures can enhance immunocompetence during early stages of infection, these same temperatures can speed the growth of the pathogen during later stages (Warwick 1991). This problem would seem most relevant when the heat tolerance of the pathogen exceeds that of the host. In such cases, we might expect infected individuals to select relatively low temperatures to slow the growth of the pathogen. Some combinations of host and pathogen do cause the host to select lower temperatures during infection. For example, snails (Lymnaea stagnalis) reduced their preferred body temperature from 25◦ C to 20◦ C when infected with either of two species of trematodes, but retained a preference for 25◦ C when infected with a third species (Zbikowska 2004). Perhaps, these different responses would make sense if we knew the thermal sensitivities of the host and parasites. Still, not all organisms reduce their body temperature during infection when the growth of the pathogen would suffer. Although tent caterpillars inoculated with a nucleopolyhedrovirus died faster at higher temperatures, inoculated and uninoculated individuals preferred relatively high temperatures, which accelerated the death of infected individuals (Frid and Myers 2002). As with fever, a reduction in body temperature during infection probably causes some loss of performance for the host. Therefore, thermoregulation during infection reﬂects a compromise between maximizing survivorship and maximizing other performances.

4.5.2 Thermoregulation during pregnancy Another excellent example of a conﬂict imposed by different thermal optima involves thermoregulation during pregnancy. Both oviparous and viviparous animals may change their thermoregulatory behavior when they carry developing offspring (Table 4.3). Beuchat and Ellner (1987) hypothesized that this behavior results from a mismatch between the thermal optima for mothers and embryos. In this case, optimizing embryonic performance enhances current reproduction whereas optimizing maternal performance enhances both current and future reproduction. Using the thermal sensitivities of maternal and embryonic performances, they modeled the optimal body temperature for pregnant lizards of a viviparous species (Sceloporus jarrovi). Beuchat and Ellner assumed the temperature during gestation was the only factor that inﬂuenced the performance of a mother and her embryos. The performance of the mother was modeled as the product of survivorship (px ) and growth (δ); growth contributed to future reproduction because litter size increased linearly with increasing body size. The performance of embryos was modeled as their probability of survival to birth. Given the litter size at age x, the probability of embryonic survival could be used to calculate the thermal sensitivity of fecundity (mx ). Empirical data from laboratory experiments were used to parameterize the thermal sensitivities of px · δ and mx in Sceloporus jarrovi (Fig. 4.17). The optimal body temperature—that is, the one that maximizes the ﬁtness of the mother—resides somewhere between the body temperature that maximizes current fecundity (mx ) and the body temperature that maximizes future fecundity (px · δ). This optimum depends on the expected variance of body temperature because deviations from the mean body temperature reduce ﬁtness asymmetrically (see Fig. 4.17). For S. jarrovi, the optimal body temperature was predicted to be 31.4◦ C under the observed variance in body temperature and 34◦ C for no variance in body temperature. The former value very nearly matches the mean body temperature of pregnant females (32◦ C). Patterns of thermoregulation during infection and pregnancy clearly illustrate the need to consider all

T H E R M O R E G U L AT I O N

107

Table 4.3 Viviparous reptiles tend to increase their body temperature when gravid, but the behaviors of oviparous reptiles vary among species. For each study, the effect of gravidity on the mean body temperature is given. Studies published prior to 1980 were reviewed by Shine (1980). Species Viviparous Chalcides ocellatus Charina bottae Crotalus viridis Crotalus viridis Hoplodactylus maculates Mabuya multifasciata Malpolon monspessulanus Nerodia rhombifera Nerodia sipedon Sceloporus bicanthalis Sceloporus grammicus Sceloporus jarrovi Thamnophis elegans Thamnophis elegans Thamnophis sirtalis Zootoca vivipara Oviparous Antaresia childreni Hemidactylus frenatus Lepidodactylus lugubris Natrix natrix Podarcis muralis Podarcis muralis Sceloporus aeneus Sceloporus gadoviae Sceloporus scalaris Sceloporus undulatus Sceloporus undulatus Sceloporus virgatus Uromastyx philbyi Urosaurus ornatus

Setting

Effect

Source

Laboratory Field Field Field Laboratory Laboratory Field Field Field Laboratory Field Laboratory Laboratory Field Field Laboratory

Higher Higher Same Higher Higher Lower Higher Higher Higher Same Lower Lower Higher Higher Higher Lower

Daut and Andrews (1993) Dorcas and Peterson (1998) Charland and Gregory (1990) Graves and Duvall (1993) Rock et al. (2000) Ji et al. (2007) Blázquez (1995) Tu and Hutchison (1994) Brown and Weatherhead (2000) Andrews et al. (1999) Andrews et al. (1997) Mathies and Andrews (1997) Gregory et al. (1999) Charland (1995) Charland (1995) Gvoždík and Castilla (2001)

Laboratory Field Field Field Field Laboratory Laboratory Field Field Laboratory Field Field Lab Field

Higher Same Same Lower Lower Same Same Same Higher and lower1 Same Same Lower Higher Same

Lourdais et al. (2008) Werner (1990) Werner (1990) Isaac and Gregory (2004) Braña (1993) Braña (1993) Andrews et al. (1999) Lemos-Espinal et al. (1997) Smith et al. (1993) Angilletta et al. (2000) Gillis (1991) Smith and Ballinger (1994) Zari (1998) Smith and Ballinger (1995)

1The effect of reproductive condition on body temperature differed between two populations.

potential beneﬁts of thermoregulation. Even when multiple performances proceed best at a single body temperature, their thermal sensitivities might differ radically (see Fig. 4.16). By considering both energetic and nonenergetic beneﬁts, Beuchat and Ellner constructed a model that predicted thermoregulation more accurately. Similar models could be developed for thermoregulation during infection or for other scenarios involving thermoregulatory conﬂicts.

4.6 Do organisms thermoregulate less precisely when the costs are greater? A major motivation for Huey and Slatkin’s model was the energetic cost of thermoregulation that arises from shuttling between preferred microclimates. Both the frequency distribution and the spatial distribution of operative temperatures determine this cost. Thermoregulation would be less costly when preferred microclimates occur more

108

T H E R M A L A D A P TAT I O N

Relative value

(a) 1.0

p1d 0.5 m2

0 26

28

30

32

34

36

38

(b)

Relative value

1.0

p1d

0.5 m2

0 26

28

30 32 34 Temperature (°C)

36

38

Figure 4.17 (a) Observed relationships between mean body temperature (SD = 0◦ C) and estimates of the current fecundity (m2 ) and the future fecundity (px · δ ) of pregnant lizards. (b) Modeled relationships between mean body temperature (SD = 1.4◦ C) and estimates of current fecundity (mx ) and future fecundity (px · δ ). Adapted from Beuchat and Ellner (1987) with permission from the Ecological Society of America.

frequently in space. Furthermore, the movement required to locate a preferred microclimate should decrease as the environment becomes patchier (i.e., negatively autocorrelated). Because of diel shifts in the angle of solar radiation, organisms must move among ephemeral patches to maintain their preferred body temperatures throughout the day. These activities are the implicit cause of the energetic cost of thermoregulation in Huey and Slatkin’s model. The optimal level of thermoregulation decreases (k → 1) as the cost of thermoregulation (c) increases. Several factors could raise the cost of

thermoregulation. As the height of the cost curve increases, the net beneﬁt of thermoregulation decreases at all environmental temperatures, which tends to favor less precise thermoregulation. A higher cost curve could result from organismal factors such as inefﬁcient locomotion or inadequate thermoperception, or from environmental factors such as the patchiness of the thermal landscape. Additionally, the optimal level of thermoregulation can decrease as operative temperatures decrease. In colder environments, the difference between the operative temperature and the thermal optimum increases; this effect would increase the cost of thermoregulation in certain environments (see Fig. 4.12a). Importantly, we must assume the spatial autocorrelation of operative temperatures does not change, even though the mean operative temperature does. The best evidence that energetic costs affect thermoregulatory behaviors comes from an experiment conducted by Withers and Campbell (1985). They constructed an experimental apparatus known as a shuttle box. The box contained an incandescent bulb at each end, which was activated when an animal broke a beam created by a pair of photocells. Once activated, the bulb would emit heat for a speciﬁed period but could not be reactivated until the animal moved to the opposite side of the box and activated the other bulb. Thus, an animal had to shuttle between the two sides of the box to maintain a body temperature that exceeded the ambient temperature. The frequency of shuttling required for thermoregulation depended on how long a bulb remained lit once it was activated. By varying this period between 30 and 180 seconds, Withers and Campbell directly manipulated the energetic cost of thermoregulation. When desert iguanas (Dipsosaurus dorsalis) were placed in the shuttle box, they behaved exactly as predicted by Huey and Slatkin’s model (Fig. 4.18). Iguanas shifted from nearly perfect thermoregulation at 180 seconds of light per activation to nearly perfect thermoconformation at 30 seconds! To this day, Withers and Campbell’s ingenious experiment remains the best demonstration of the impact of energetic costs on thermoregulation. When assessing the role of energetic costs in generating diel, seasonal, and geographic variation

T H E R M O R E G U L AT I O N

109

(a) 70 Operative temperature (°C)

36

32

28

30 s 45 s 90 s 180 s

24 20

25 Air temperature (°C)

30

Figure 4.18 A lower energetic cost of thermoregulation led to a greater degree of thermoregulation by desert iguanas. Error bars represent 2 standard errors. Body temperatures of desert iguanas were measured in an arena that required shuttling at certain intervals to maintain a radiative heat source. The 30 second interval demanded the most energetic expenditure and the 180 second interval required the least. The dashed line represents the hypothetical slope of a thermoconformer. When placed in a thermal gradient, where no shuttling was required for thermoregulation, the same iguanas preferred 38.7±0.4◦ C. Data are from Withers and Campbell (1985).

in body temperature, a major problem emerges: researchers rarely record sufﬁcient data to estimate the beneﬁts and costs of thermoregulation. In most studies to date, operative temperatures were sampled haphazardly or selectively (see Chapter 2), precluding an accurate estimate of the thermal quality of the environment. Furthermore, researchers rarely present the frequency distribution of operative temperatures, preventing us from teasing apart energetic costs and thermal constraints. The following example illustrates this problem. Grant (1990) compared the thermoregulation of lizards (Sceloporus merriami) on two sides of a canyon in southwest Texas. In the morning (1000–1200 hours), the thermal optimum for energy intake was readily attainable on the east side of the canyon, but was rarely attainable on the west side (Fig. 4.19). As predicted by the theory, lizards on the east side thermoregulated more precisely during this period of the day. However, lizards on both sides of the canyon were unable to attain their thermal optimum during

60 50 40 30 20 0600 0800 1000 1200 1400 1600 1800 2000 2200

(b) 70 Operative temperature (°C)

Body temperature (°C)

40

60 50 40 30 20 0600 0800 1000 1200 1400 1600 1800 2000 2200 Time of day (h)

Figure 4.19 The diel patterns of operative temperatures for Sceloporus merriami differed dramatically between the (a) east and (b) west sides of a canyon in Big Bend National Park, Texas, USA. In each plot, the solid lines denote minimal, third quartile, median, ﬁrst quartile, and maximal operative temperatures. The dashed line marks the thermal optimum for energy gain (Beaupre et al. 1993). During the majority of the day, a lizard could not attain this thermal optimum on either side of the canyon. Adapted from Grant (1990) with permission from the Ecological Society of America.

most of the day. Without this detailed knowledge of the relative abundance of operative temperatures over space and time, we could easily draw a false conclusion about thermoregulation. For lack of better information, we might assume the cost of thermoregulation increases as preferred microclimates become rare. Indeed, thermoregulation in forested and open habitats have been compared, under the assumption that thermoregulation costs more energy in forested habitats because basking sites are less abundant (see, for example, Hertz 1974; Huey 1974). Similarly, body temperatures of a species have been compared along latitudinal

T H E R M A L A D A P TAT I O N

(a) 10 Accuracy of thermoregulation, db (°C)

and altitudinal clines. The mean body temperature decreases with increasing latitude or altitude in some species, (Hertz 1981, 1983; Kiefer et al. 2005; Pianka 1970; Van Damme et al. 1989), but does not do so in other species (Angilletta 2001; Hertz and Huey 1981). Because these intraspeciﬁc comparisons were not coupled with random sampling of operative temperatures, we have no way of really knowing whether colder environments imposed greater energetic costs of thermoregulation. By considering frequency distributions of operative temperatures, Blouin-Demers and Nadeau (2005) provided the ﬁrst broad and rigorous evaluation of the impact of energetic costs on thermoregulation in natural environments. They reasoned that the index de provided a good measure of the cost of thermoregulation because a large de means the environment is (on average) much hotter or colder than the organism prefers (see Box 4.2). This reasoning makes sense if we envision a costly environment as one in which an animal must exert greater effort to ﬁnd preferred microclimates. If fewer sites lie within the preferred range of temperatures, these sites should be harder to locate (and might also be farther from other important resources, such as food and refuge). Assuming de reﬂects the increase sharply relative cost of thermoregulation, we should expect the accuracy of thermoregulation (db ) to increase sharply as de increases (i.e., db > de ; Fig. 4.20a). Plotting independent contrasts for 22 species of lizards, Blouin-Demers and Nadeau observed just the opposite (Fig. 4.20b). Moreover, Row and BlouinDemers (2006) discovered that similar plots of db versus de for individual snakes (Lampropeltis triangulum) also differed from the predicted relationship. If Withers and Campbell’s experiment provided such strong evidence that energetic costs inﬂuence thermoregulation, why do thermoregulatory behaviors in natural environments seem to contradict this conclusion? Blouin-Demers and Nadeau hypothesized that individuals cannot afford to thermoconform when operative temperatures differ greatly from their preferred temperature because the loss of beneﬁt associated with thermoconformation exceeds the greater cost of thermoregulation. This idea has some merit but cannot be validated without accurate measures of beneﬁt and cost curves in the relevant environments. Additional possibilities

Thermoconformation (m = 1)

8 Huey–Slatkin model (m > 1)

6

4 d e – db 2 Opposite of the Huey– Slatkin model (m < 1)

0 0

2

4

6

8

10

(b) Accuracy of thermoregulation, db (°C)

110

10

Thermoconformation (m = 1)

8 6 4 2 0 0

2 4 6 8 10 Thermal quality of habitat, de (°C)

Figure 4.20 (a) Using the thermal quality of the environment (de ) to estimate of the cost of thermoregulation, Blouin-Demers and Nadeau predicted the slope (m) of the relationship between de and the accuracy of thermoregulation (db ). The bold line depicts the relationship for a perfect thermoconformer (m = 1), whereas the dashed lines depict relationships presumably consistent (m > 1) and presumably inconsistent (m < 1) with Huey and Slatkin’s model. (b) The observed relationship between de and db opposes the predicted relationship. Adapted from Blouin-Demers and Nadeau (2005) with permission from the Ecological Society of America.

become clear in light of the assumptions made by these investigators. First, they assumed the thermal sensitivity of energy gain was the same for all animals. Second, they assumed preferred body temperature equaled the thermal optimum for energy gain. Third, they assumed de adequately captured the energetic cost of thermoregulation. If any of these assumptions proves invalid, we should question the inference drawn from the comparative data.

T H E R M O R E G U L AT I O N

Finally, both the optimality model and the comparative study ignore nonenergetic costs of thermoregulation. Costs of predation risk and missed opportunities also inﬂuence the optimal thermoregulation and these costs have not been integrated with the energetic cost. Even if we attribute a paramount signiﬁcance to energetic costs, we will have a difﬁcult time constructing a cost curve from measurements of de because this index ignores the spatial structure of the environment. After all, de tells us only the mean difference between the operative environmental temperature and the preferred body temperature. This index fails to capture the statistical variance or spatial distribution of operative temperatures. Without controlling for environmental variance and spatial structure, the use of de to estimate energetic costs can lead to erroneous conclusions (Sears 2006). To comprehend the signiﬁcance of spatial structure, picture an environment with a frequency distribution of operative temperatures, characterized by some value of de . This statistical distribution (and hence the de ) tells us nothing about the spatial distribution of these temperatures. In fact, we can imagine many spatial distributions for any statistical distribution. A clumped distribution of operative temperatures poses a different challenge to an organism than does a dispersed distribution. When favorable temperatures are clumped, other resources would be either more or less likely to occur near preferred microclimates. Moreover, clumped distributions caused by large clearings or canopies should change more predictably as the sun moves across the sky. Yet, clumped and dispersed distributions can yield identical values of de ! Because the cost of thermoregulation depends on both the spatial and statistical distributions of operative temperatures, one must control for spatial structure when testing Huey and Slatkin’s model. Using computer simulations, we can estimate cost curves for speciﬁc thermal environments (Box 4.4); however, applying this approach requires extensive knowledge of the biology of the organism and the structure of the environment. Simply put, estimating the cost of thermoregulation is a complex task. This task becomes even more complex when we add the nonenergetic costs.

111

4.7 Nonenergetic costs of thermoregulation 4.7.1 Aggressive interactions with competitors In a seminal paper, Magnuson and colleagues (1979) argued convincingly that competition for temperature could be viewed in the same way as competition for food. After all, preferred microclimates occur in space, and ecologists have long known that organisms compete directly for space (Connell 1961). Magnuson and colleagues characterized the thermal niche by statistically describing the temperatures selected in a thermal gradient (e.g., the median or range ± a percentage of the distribution). By adopting this view, we could characterize the use of available space in relation to its thermal quality (Huey 1991; Tracy and Christian 1986). When two species compete for thermal resources, the thermal niche of the subordinate competitor or both competitors could be altered by the interaction (Buckley and Roughgarden 2005). Magnuson and his colleagues (Beitinger and Magnuson 1975; Medvick et al. 1981) experimentally veriﬁed the impact of intraspeciﬁc competition on thermoregulation. In a series of experiments, bluegill sunﬁsh (Lepomis macrochirus) were placed individually or socially in a tank with two compartments: an optimal habitat (31◦ C) and a suboptimal habitat (a constant temperature between 21◦ C and 29.5◦ C). In isolation, ﬁsh preferred the optimal habitat over the suboptimal one (Fig. 4.23a). In pairs, the dominant ﬁsh always occupied the optimal habitat and the subordinate did not (Fig. 4.23b). In larger groups, all ﬁsh crowded into the optimal habitat, particularly when the suboptimal habitat was much colder than the optimal habitat (Fig. 4.23c–f). Observations of reptiles also support the view that intraspeciﬁc competition affects thermoregulatory behavior. According to Seebacher and Grigg (2000), crocodiles often emerge from water to bask, but dominant crocodiles chase subordinate crocodiles back into the water before these subordinates can reach their preferred body temperature. Moreover, subordinates do not return to bask as frequently after being chased. Using mathematical models, the body temperatures of subordinate crocodiles

112

T H E R M A L A D A P TAT I O N

were estimated to have been 6◦ C below the body temperatures of dominant crocodiles. Competition could cause rapidly reversible or developmentally ﬁxed changes in thermoregulatory behavior. For example, Regal (1971) noticed a marked change in basking when a male lizard was exposed to another male and a rapid reversal of behavior when the competitor was removed. In a more rigorous study of the same phenomenon, lizards of a dominant species (Podarcis sicula) used warm microclimates more frequently when sharing an enclosure with a subordinate competitor (P. melisellensis) than when occupying a similar enclosure alone (Downes and Bauwens 2002). In contrast, individuals can express combinations of social and thermoregulatory behaviors in the absence of competitors; dominant, orange-colored males of the lizard Pseudemoia entrecasteauxii preferred a slightly higher mean temperature than subordinate,

white-colored males, even when placed individually in a thermal gradient (Stapley 2006). Such combinations of social and thermoregulatory strategies could develop during early encounters with conspeciﬁcs and remain relatively ﬁxed. Differences in thermoregulatory behavior between competitors, whether ﬂexible or ﬁxed, could minimize risks of stress, injury, or death.

4.7.2 Risk of predation or parasitism When ectotherms thermoregulate, they often seek out areas that receive intense solar radiation. This behavior would place them in open habitats, where they would draw the attention of predators (Webb and Whiting 2005). Just about every environment offers some form of refuge from predation, but

Box 4.4 Estimating cost curves by computer simulation Estimating the energetic cost of thermoregulation seems hopelessly difﬁcult because this cost depends on the statistical and spatial distributions of operative temperatures, as well as physical and physiological properties of the organism. Still, we can estimate this cost through computer simulations of movement in virtual landscapes. Using such simulations, Sears and Angilletta (2007) calculated cost curves for hypothetical thermal

(b)

(c)

Frequency

(a)

environments. Simulated individuals were assigned properties of a small lizard (≈ 10 g), such as Sceloporus undulatus (Angilletta 2001; Dzialowski and O’Connor 2004). These individuals moved through a 60 m × 60 m grid, with each cell (1 m2 ) having an operative temperature drawn from a bimodal distribution (Fig. 4.21a). Holding this statistical distribution constant, we conﬁgured the thermal landscape in various ways; for example, Fig. 4.21 shows

60 50 40 30

50 60 30 40 Operative temperative (°C) Figure 4.21 (a) A bimodal distribution of operative temperatures was used to simulate the energetic cost of behavioral thermoregulation. (b and c) Depending on the simulation, these temperatures were distributed in one large clump or in many small patches. See plate 2.

continues

T H E R M O R E G U L AT I O N

113

Box 4.4 Continued predicting thermoregulation. Finally, the explicit mapping of operative temperatures enables us to consider spatial covariations between temperature and other factors, such as food, water, and predators. These spatial covariations ultimately determine the costs of predation risk and missed opportunities.

1000 Clumped distribution

800 600 400 Distance moved (m)

a clumped spatial distribution and a dispersed spatial distribution. For each spatial distribution, we simulated behavioral thermoregulation and estimated its cost. Energetic costs of thermoregulation were estimated as the movement required to achieve a preferred body temperature. At the start of a simulation, an individual was placed in a randomly selected microenvironment and was assigned a body temperature equal to the operative temperature. From this initial position, the individual moved until its body temperature entered the set-point range (33−36◦ C). The angle and distance of each movement were randomly drawn from uniform distributions (angle = 0−360◦ ; distance = 0−3 m). The change in body temperature at each interval was calculated from the thermal time constant of the individual (5 minutes) and the operative temperature of the microenvironment (Gates 1980). The distance moved during the simulation provides a simple proxy for the energetic cost of thermoregulation. By replicating this simulation 1000 times, we obtained costs of thermoregulation for a range of initial operative temperatures. The shape of the cost curve, as conceived by Huey and Slatkin (1976), was inferred by plotting the distance moved against the absolute deviation between the thermal optimum and the initial operative temperature (|Topt − Te |). The spatial structure of the environment had a dramatic impact on the cost of thermoregulation (Fig. 4.22). In the clumped environment, more movement was necessary to locate a preferred microenvironment when the initial operative temperature deviated further from the set-point range. In the patchy environment, the initial operative temperature had little effect on the movement required to thermoregulate. These results make sense considering the spatial distributions used in the simulation. In a clumped environment, great distances might be required to reach a preferred microclimate if the initial microclimate deviates greatly from the preference (see Fig. 4.21b). In a patchy environment, an individual always resides near a preferred microclimate regardless of its position (see Fig.4.21c). On average, the energetic cost of thermoregulation in a clumped environment greatly exceeds that in a patchy environment. This approach to estimating the cost of thermoregulation confers several advantages. First, we can consider properties of animals, such as perception, mobility, and memory. Second, we can account for both spatial and statistical distributions of operative temperatures when

200 0 0

2

4

6

8

10

12

14

10

12

14

1000 Dispersed distribution

800 600 400 200 0 0

2

4

6 8 |Topt – Te|

Figure 4.22 The energetic cost of behavioral thermoregulation depends on the spatial structure of the environment. In an environment where preferred microclimates occur in one large clump (positive spatial autocorrelation), simulated individuals had to move greater distances to thermoregulate as the operative temperature of their initial microenvironment (Te ) deviated further from the optimal temperature range. In an environment where preferred microclimates were dispersed (negative spatial autocorrelation), simulated individuals had to move only small distances to thermoregulate, regardless of their initial microenvironment. For calculations of |Topt − Te |, the thermal optimum (Topt ) was represented by either the lower or upper limit of the set-point range (33◦ C or 36◦ C).

114

T H E R M A L A D A P TAT I O N

(a) 100

(b) 100

Median % of fish in the optimal habitat

N=2 N=1

50

50 0

0 (c) 100

(d) 100 N=4

50 0

0

(e) 100

(f) 100

50 0

N = 16 0

N=8

50

50

N = 32

0 5 10 0 Thermal difference between habitats (°C)

5

10

Figure 4.23 The percentage of ﬁsh using an optimal habitat instead of a suboptimal habitat depended on the number of competing ﬁsh (N) and the thermal difference between habitats. For any particular trial, a ﬁsh was either alone (a) or in the presence of some number of conspeciﬁcs (b–f). Competition forced some ﬁsh into the suboptimal habitat, except in cases when the difference in temperature between habitats was extreme. Adapted from Medvick et al. (1981) with permission from the American Society of Ichthyologists and Herpetologists.

refuges are generally insulated and cool. Consequently, a fundamental tradeoff exists between the regulation of temperature and the risk of predation. Recent modeling has shed light on the ways in which predation risk can inﬂuence thermoregulation in terrestrial and aquatic environments. For terrestrial ectotherms, we shall consider the optimal use of burrows or crevices during thermoregulation. For aquatic ectotherms, we shall consider the optimal migration between deep, cold waters and shallow, warm waters. Imagine a terrestrial animal that basks to maintain its thermal optimum for performance, but this behavior attracts the attention of a predator. The animal can avoid predation by ﬂeeing to a refuge; however, its body temperature will fall below its thermal optimum within the refuge. When should the animal emerge from the refuge to continue basking? To determine the optimal duration to spend within a refuge, one must deﬁne the costs and beneﬁts of abandoning thermoregulation. Obviously, the beneﬁt of staying in the refuge equals the increase in survivorship from avoiding the predator. The cost of

staying in the refuge equals the loss of performance caused by the drop in body temperature (Amo et al. 2007a,b). Assuming the beneﬁt and the cost can be expressed in the same currency (ﬁtness), we can ﬁnd the optimal emergence time. Polo and colleagues (2005) modeled this problem with speciﬁc cost and beneﬁt curves. They assumed attack by a predator imposes a risk of predation (R), but this risk decreases gradually over time (x): 2 R(x) = x/a , e + e−x/a

(4.9)

where a determines the rate at which the risk of predation decays over time. Although a refuge offers protection from predation, the cost of remaining in the refuge increases asymptotically over time: C(x) = m(1 − e−x/n ),

(4.10)

where m and n deﬁne the shape of the cost curve. Based on this model, the time spent in a refuge

T H E R M O R E G U L AT I O N

PR Chigh Ehigh

Clow

Elow

Cost of refuge use (C)

Predation risk (PR)

(a)

Emergence time

C1

C2 C 3 C 4

E4 E3 E2 PR4 E1

PR3 PR1

Cost of refuge use (C)

Predation risk (PR)

(b)

PR2 Emergence time

Figure 4.24 An optimality model predicts the time required to emerge from a cold refuge after an attack by a predator. (a) The optimal decision (E ) depends on the change in predation risk over time (PR) and the energetic cost of remaining within the retreat (C ). Arrows depict the optimal time of emergence for retreats imposing high and low costs. (b) The optimal decision changes with subsequent attacks if we assume an organism’s predation risk increases or its energetic cost decreases with each new attack. Adapted from Polo et al. (2005) with permission from Evolutionary Ecology Ltd.

should decrease as the temperature of the refuge decreases (Fig. 4.24a). This result occurs because the loss of performance when hiding in a cold refuge exceeds the loss of performance when hiding in a warm refuge. Polo and colleagues used their model to predict the behavior of alpine lizards in an artiﬁcial thermal arena. They set up a linear gradient of basking sites and placed a refuge in the coolest region. By adjusting the height of the heat sources, they could control the operative temperatures throughout the gradient. The operative temperature of the warmest region lay within the lizard’s range of preferred body temperatures (31–34◦ C), whereas the operative temperature in the refuge equaled either 27◦ C (low-cost refuge) or 18◦ C (high-cost refuge). After allowing a

115

lizard to habituate to the gradient, it was attacked by a simulated predator (a paint brush). As predicted, lizards provided with the warm refuge took about three times as long to emerge as did lizards provided with the cool refuge. The researchers continued to attack each lizard after the initial emergence and noted that emergence time increased by about 1 minute with each successive attack. They conjectured this pattern reﬂects changes in both the perceived risk and the cost curve with repeated attacks. If each attack increases a lizard’s perceived risk of predation, the risk curve, R(x), becomes shallower during successive attacks (however, predation risk would actually increase with each attack if lizards tire progressively during repeated escapes). Additionally, a lizard would be cooler than it prefers if attacked when emerging from the refuge. Because the initial body temperature would be lower at the onset of the second attack, the lizard would experience a different cost curve, C(x). The combined changes in these two curves should favor longer emergence times with repeated attacks (Fig. 4.24b). Results of a ﬁeld experiment with the same species were fairly consistent with the model (Martin and Lopez 2001). Furthermore, the model also explained differences in the use of refuges between populations; lizards in a safer environment waited longer to ﬂee and emerged sooner after hiding than did lizards in a risky environment (Diego-Rasilla 2003). Predation risk during basking can cause ineffective thermoregulation (Herczeg et al. 2008), which would ultimately slow growth and reproduction (Downes 2001). The risk of predation also affects the thermoregulatory behaviors of aquatic organisms. Many species of zooplankton migrate vertically on a daily basis, spending days in cool, deep waters and nights in warm, shallow waters. Several lines of evidence suggest both energetic cost and predation risk modulate the degree of vertical migration (reviewed by Lampert 1989). The intensity of daily migration by water ﬂeas (Daphnia magna) depended strongly on the concentration of ﬁsh kairomones (Loose and Dawidowicz 1994). In the presence of concentrated kairomones, individuals migrated to cool, deep water during the day. In the presence of dilute or no kairomones, the water ﬂeas remained in warm

116

T H E R M A L A D A P TAT I O N

surface waters at all times. Predation risk also seems to explain differences in the direction of migration (Lampert 1989); for example, an exceptional species, which spends the day in shallow waters and the night in deep waters, avoids predation by a larger species of zooplankton (Ohman et al. 1983). Similarly, snails avoided areas of a thermal gradient that were concentrated with the odor of crushed conspeciﬁcs. Upon encountering these areas, an individual withdrew immediately and often moved as far as possible from the source of the cue (Gerald and Spezzano 2005). Although snails usually favor very low temperatures, their movements in the presence of conspeciﬁc cues resulted in higher body temperatures. Ectotherms may also alter their use of microclimates in response to the risk of parasitism. For example, certain ticks attach to animals as they pass through vegetation. Kerr and Bull (2006) argued this localized risk of parasitism affects the use of microenvironments by the sleepy lizard (Tiliqua rugosa). Lizards had the greatest risk of acquiring ticks when seeking thermal refuge in bushes. The risk of parasitism doubled or tripled as temperatures rose within bushes. Because the most extreme thermal conditions enhance both the need for a thermal refuge and the risk within that refuge, a tradeoff between thermoregulation and parasitism exists. As with the risk of predation, the risk of parasitism could impose a cost of thermoregulation that deserves the attention of both theorists and empiricists.

4.7.3 Risk of desiccation Although evaporative water loss enables organisms to avoid overheating, dehydration causes a loss of performance that can ultimately lead to death. For this reason, we should expect the hydration of an organism to inﬂuence its thermoregulatory strategy. Since evaporation proceeds more rapidly at higher temperatures, we could reasonably predict that poorly hydrated organisms would behaviorally avoid high temperatures. Studies of animals and plants support this prediction. Dehydrated animals typically prefer lower temperatures than do hydrated animals (Crowley 1987; Dohm et al. 2001; Ladyman and Bradshaw 2003; Smith et al. 1999).

Among 12 species of Australian frogs, preferred body temperature correlated positively with the water resistance of skin (Tracy and Christian 2005). In plants, the need for solar radiation inextricably links thermoregulation and photosynthesis. Plants that track the sun to absorb visible light must also use evaporation to avoid overheating (Ehleringer and Forseth 1980; Galen 2006). At the wilting point, leaves and ﬂowers abandon solar tracking to avoid the resulting damage caused by high temperatures or poor hydration (Forseth and Ehleringer 1980). In natural environments, a shortage of water alters the pattern of activity (Lorenzon et al. 1999) and the use of microhabitats (Cohen and Alford 1996) in ways that affect the body temperatures of animals.

4.7.4 Missed opportunities for feeding or reproduction Organisms do more than thermoregulate; they must also ﬁnd food, defend territories, and attract mates. When thermoregulation conﬂicts with these other activities, an individual incurs a cost of missed opportunities. Because activities such as ﬁnding and handling prey determine the energy gain, the beneﬁt of thermoregulation depends on the degree to which thermoregulatory behaviors cause missed opportunities. For example, a conﬂict between food and temperature drives the daily vertical migration of dogﬁsh (Scyliorhinus canicula). In an artiﬁcial thermal gradient, these ﬁsh preferred cool water. Yet individuals in their natural environment spent the night within cool, deep waters and the day within warm, shallow waters, because shallower waters provided far greater densities of food (Sims et al. 2006). In some lakes, Daphnia can ﬁnd more food in deep water, but grow more slowly there because of low temperatures. By manipulating vertical clines in artiﬁcial water towers, Kessler and Lampert (2004) showed that Daphnia migrate vertically in the water column to optimize the combination of food and temperature (Fig. 4.25). When presented with a steep thermal gradient, Daphnia spent most of their time at intermediate depths. As thermal gradient dampened, Daphnia spent more time in deep waters. Bluegill sunﬁsh also chose a habitat for both its thermal properties and its prey density (Wildhaber 2001). When preferred temperatures and preferred

T H E R M O R E G U L AT I O N

Steep gradient

Moderate gradient

117

Mild gradient

0 Day

Night

Day

Night

Day

Night

1

Depth (m)

2 3 4 5 6

10°C

5°C

2°C

7 50

25

25

50

50 25 25 50 Percentage of the population

50

25

25

50

Figure 4.25 The vertical distribution of water ﬂeas shifted from shallow water to deep water as the vertical gradient in temperature became milder. The horizontal bars in each plot show the mean percentage of the population (±SE) dwelling at the respective depths during the day (unﬁlled bars) and night (ﬁlled bars). The temperature of the top layer of water was always 20◦ C. The temperature differences between the top and bottom layers of water are given in the panels. The dashed line indicates the thermocline (the point at which the temperature changes rapidly with depth). Adapted from Kessler and Lampert (2004) with kind permission from Springer Science and Business Media. © Springer-Verlag 2004.

foods occur in different places, perfect thermoregulation can even decrease energy gain. Thus, we must consider both thermal and nutritional resources when predicting the use of microhabitats (e.g., Crowder and Magnuson 1983; Riechert and Tracy 1975; Wildhaber and Lamberson 2004). Gvoždík (2002) investigated the tradeoff between basking and foraging for lizards (Zootoca vivipara) in four environments along an altitudinal gradient of 250–1450 m. A thermoregulator was assumed to maintain its temperature between lower and upper set points (Tmin and Tmax , respectively). Set points were estimated from body temperatures selected in an artiﬁcial thermal gradient (see Box 4.1). A lizard was hypothesized to initiate basking when its temperature fell to the lower set point and cease basking when its temperature rose to the upper set point. Using the biophysical model of Bakken and Gates (1975), Gvoždík estimated the duration of basking (tbask ) required to thermoregulate: Tmax − Te tbask = −τh ln , (4.11) Tmin − Te where τ h equals the time constant for heating. Once an animal reached its upper set point, it could forage until its temperatures dropped to the lower set point.

Thus, the time available for foraging (tforage ) was as follows: Tmin − Te , (4.12) tforage = −τc ln Tmax − Te where τc equals the time constant for cooling. To quantify the tradeoff between basking and foraging, Gvoždík assumed lizards cannot bask and forage simultaneously. Based on operative temperatures measured on clear days, he estimated the time required for basking at each elevation. At the highest elevation (1450 m), operative temperatures were mainly below the set-point range. Consequently, lizards at this elevation would have needed to spend about 50% of their time basking to maintain their temperature within the set-point range. Among the four sites, the estimated proportion of time required for basking was very highly correlated with the observed proportion of time spent basking. This greater duration of basking explains how lizards at high elevation were able to thermoregulate as accurately and effectively as lizards at lower elevations. If Gvoždík’s assumptions were correct, lizards at the highest elevation might have gained less energy than lizards at lower elevations, despite their effective thermoregulation.

118

T H E R M A L A D A P TAT I O N

Opportunities for courtship and mating might also be compromised by thermoregulation. Males cannot bask uninterrupted while defending their territories. Long bouts of courtship or mating cause body temperatures of the participants to rise above or fall below the preferred range (Shine et al. 2000). Regulating conductive or convective exchange of heat can also interfere with the defense of resources and mates. Such tradeoffs between mating and thermoregulation were recently identiﬁed in male bees (Roberts 2005). Males of the solitary desert bee (Centris pallida) aggregate by the thousands in spring, attempting to mate with virgin females as they emerge from their nest. Bees either patrol the ground searching for mates or defend small territories by hovering along the edge of the aggregation. Like many insects, these bees alter the frequency of wing-beating and the ﬂow of hemolymph to thermoregulate during ﬂight. Despite these regulatory mechanisms, convective heat loss depends strongly on the speciﬁc manner of ﬂight. Patrolling males beat their wings as rapidly as hovering males, yet patrollers were 1–3◦ C cooler than hoverers. Based on these observations, Roberts concluded that patrolling males suffered greater heat loss to convection than hovering males did. Because the reproductive beneﬁts of these two strategies likely differ, a tradeoff exists between thermoregulation and mating. Although thermal biologists will need to investigate the generality of missed opportunities during thermoregulation, I suspect most organisms must choose between thermoregulation and competing activities. Given a cost of missed opportunities, individuals should endeavor to spend as little time as possible to reach their preferred body temperatures. Belliure and Carrascal (2002) carried out an experimental test of this hypothesis. They placed lizards in an enclosure with two basking sites: one site that offered mainly radiative heat exchange and another site that offered mainly conductive heat exchange. Both sites provided the same operative temperature but differed in the rates of heating they enabled. Amazingly, lizards were able to detect the difference between these sites and selected the one that offered the faster rate of heating! This result conﬁrms that, at least in one species, individuals strive to minimize missed opportunities during thermoregulation.

4.7.5 Interactions between different costs Estimating the ﬁtness consequence of a thermoregulatory strategy becomes further complicated when various costs interact with one another. For example, we know that animals avoid preferred microclimates when they occur in risky habitats (see Section 4.7.2). Although predation risk clearly deters individuals from making certain thermoregulatory decisions, the decision to avoid this risk depends on other factors, such as the density of conspeciﬁcs. Two experiments by Downes nicely illustrate this point. In the ﬁrst experiment, Downes and Shine (1998) showed that antagonistic interactions between conspeciﬁcs could alter thermoregulation under the risk of predation. Geckos were placed in enclosures that offered two retreats that differed in temperature. Solitary geckos preferred to rest in the warm retreat, but shifted their preference to the cold retreat when the scent of a predator was applied to the warm retreat. The result differed when geckos were paired with larger conspeciﬁcs; the smaller member of the pair was forced to use the warm retreat even when this site had been marked by a predator (Fig. 4.26). In the second experiment, Downes and Hoefer (2004) showed that interactions among conspeciﬁcs can sometimes reduce the cost of thermoregulation imposed by predators. When alone or in small groups, sun lizards basked less in the presence of chemical cues from snakes. In larger groups, however, the same lizards basked frequently with or without cues from the predator. When animals ﬁnd safety in numbers, the presence of many conspeciﬁcs dilutes the risk of predation during basking. Dynamic optimality models can make predictions about thermoregulation when temperature, food, and risk vary among microenvironments (Box 4.5). Such models can easily include complex biological details and are best tailored for speciﬁc organisms. If these models include an explicit consideration of spatial structure, they could not only provide more precise estimates of energetic costs but could also incorporate the spatial covariation between temperature and the factors that inﬂuence nonenergetic costs. The distribution and abundance of other resources in space almost certainly determine costs of thermoregulation. Whether an organism views

T H E R M O R E G U L AT I O N

(a)

*

*

30 25

*

20

Number of geckos in the warm retreat

15 10 5 0 (b)

12 9

*

*

15

*

to maintain relatively high and stable temperatures (Clarke and Rothery 2008), they undoubtedly pay extraordinary energetic costs of thermoregulation. Indeed, the sheer energetic extravagance of endothermy probably explains why biologists have been obsessed with its explanation (Koteja 2004). Despite considerable efforts, the evolutionary origin of endothermy remains as mysterious today as it was several decades ago. Biologists have a much better understanding of the evolutionary maintenance of endothermy in extant populations. In this section, we shall explore both the origin and maintenance of endothermy. In doing so, we shall see how Huey and Slatkin’s model of behavioral thermoregulation applies perfectly well to endothermic thermoregulation.

*

4.8.1 The evolutionary origins of endothermy

6 3 0

119

Large Small

Large Small

Large Small

First retreat

Final retreat Times entered

Figure 4.26 Selection of retreats by (a) solitary geckos and (b) paired geckos depended on the risk of predation. In each case, geckos were exposed to two conditions: a choice between cold and warm retreats without chemical cues of a predator (ﬁlled bars) and a choice between a cold retreat without chemical cues and a warm retreat with chemical cues (open bars). The dotted line shows the expected number of geckos in each retreat if no preference existed. An asterisk above a pair of bars indicates a signiﬁcant difference in choice between the two conditions (P < 0.05). Adapted from Downes and Shine (1998) with permission from Elsevier. © 1998 Elsevier Ltd.

a microclimate as favorable or not depends on its proximity to competitors, predators, parasites, prey, water, and refuge. A spatially explicit model would enable us to determine how these factors interact with temperature to affect the optimal strategy of thermoregulation.

4.8 Endothermic thermoregulation Endothermy—the regulation of body temperature via metabolic heat—occurs in mammals and birds, as well as certain insects, ﬁsh, and plants. Because many endotherms produce enough heat

Two kinds of explanations for the evolution of endothermy have been offered: a response to direct selection for thermoregulatory capacity and an indirect response to selection for metabolic capacity (reviewed by Hayes and Garland 1995; Kemp 2006). Direct selection for thermoregulatory capacity would have required the beneﬁt of thermoregulation to offset the cost of thermogenesis. By now, the beneﬁts of thermoregulation should be obvious. Besides the physiological beneﬁts of being hotter (Section 4.2.1), an organism might expand its temporal or spatial range of activity by offsetting heat loss with heat production (Crompton et al. 1978). Nonetheless, endotherms also require morphological and physiological mechanisms to store the heat produced by metabolism. Without ample size or insulation, the thermoregulatory beneﬁts of endothermy lie beyond reach. For example, even a quadrupling of metabolic rate could not increase the body temperature of a large lizard by 1◦ C (Bennett et al. 2000). Without certain morphological features, the energetic cost of elevating body temperature seems too high. Recently, Farmer (2000) proposed a twist on the original hypothesis for the evolution of endothermy. She proposed endothermy evolved because metabolic heat enables parents to warm their developing embryos. In this sense, she viewed

120

T H E R M A L A D A P TAT I O N

Box 4.5 Modeling optimal vertical movements in aquatic ectotherms

Equal risk

Risky in surface Risky in surface water and shallow waters

1.0 0.8 0.6 0.4 0.2 0

1.0 0.8 0.6 0.4

Deep

Shallow

Surface

Deep

Shallow

Surface

Deep

0

Shallow

0.2 Surface

(4.13)

Proportion in zone (day)

L = K [F (T , L, t ) − C (T , L, t )] .

Because Alonzo and Mangel assumed that fecundity depended directly on body length, the use of warm water to speed growth would enhance reproduction. Empirical estimates of performance curves were used to choose reasonable rates of feeding and metabolism over a range of body sizes. The presence of predators imposed a cost of using warm waters. Both baseline and size-speciﬁc probabilities of survivorship were considered. When deﬁning the size-speciﬁc probability, predators were assumed to prefer larger individuals over smaller ones. Both baseline and size-speciﬁc probabilities could vary among vertical zones. Empirical estimates of age distributions in natural environments were used to choose reasonable survivorships over a range of body sizes. To calculate the optimal behavior, Alonzo and Mangel had to choose a measure of ﬁtness. Fecundity and survivorship were used to calculate the reproductive value—the sum of current and expected future reproduction. The optimal strategy maximizes the reproductive value of an individual at birth, thus maximizing the expected number of offspring produced during its lifetime. Because growth and survivorship depended on size, which changed over time, Alonzo and Mangel identiﬁed the optimal behavior through dynamic programming. This modeling approach enables one to track the inﬂuence of state variables (e.g., size) on the optimal strategy over the course of an individual’s life (Mangel and Clark 1988). The use of warm, surface waters depended strongly on the level of predation (Fig. 4.27). Without size-speciﬁc

predation, the optimal behavior always involved spending the night on the surface. The optimal location during the day depended on the relative risk of predation. When the risk of predation was equal among zones, the optimal forager spent most of its day in surface water. When predation risk was greatest in surface water, the optimal forager spent most of its day in shallow water. Finally, when predation risk was great in both surface and shallow waters, the optimal forager spent most of its day in deep water. More complex patterns of optimal behavior were predicted when the risk of predation depended on size; speciﬁcally, optimal foragers split their diurnal activity between surface water and deeper waters when predation risk was concentrated. Although these qualitative predictions seem obvious in hindsight, Alonzo and Mangel’s model provides a framework for generating quantitative predictions under many scenarios. Biologists could parameterize the model for other aquatic species or develop similar models for terrestrial species. Dynamic programming offers an efﬁcient means of calculating the optimal thermoregulation under a variety of constraints.

Proportion in zone (night)

Alonzo and Mangel (2001) modeled the activity and growth of krill along a vertical cline in water temperature. The environment was structured into zones, comprising surface, shallow, and deep waters. Abundant prey resided in warm, surface water, but refuge from predators occurred in cool, deep water. Krill could adopt one of three behaviors: (1) spend both night and day at the surface, (2) spend the night at the surface but migrate to shallow or deep water during the day, or (3) spend both night and day in shallow or deep water. The energetic consequence of this decision was based on the thermal sensitivity of growth rate. Because the energetics of growth depended on body size, a complex function related food consumption (F ) and metabolic costs (C ) to the time of year (t ), the length of the animal (L), and the temperature of its vertical zone (T ). By adding a constant (K ) that converts energy into a change in length, they could calculate the growth of an individual for each zone:

Figure 4.27 Dynamic optimization predicts that vertical migration by krill will depend on the risk of predation. The proportion of krill predicted in each vertical zone is shown for three scenarios of predation risk. Krill should spend their nights and days in warm surface water when equal predation risk exists at all depths. Concentrated predation risk in shallow or surface waters favors a shift to deeper water by day. Adapted from Alonzo and Mangel (2001) with permission from the Ecological Society of America.

T H E R M O R E G U L AT I O N

endothermy as a form of parental care. A higher temperature would enable faster development and possibly cause greater survival of offspring. From the parent’s perspective, the energetic cost of reproduction would increase, but the rapid development of offspring would provide additional time to acquire food or reproduce further. By quantifying the beneﬁts of endothermic parental care and weighing them against the costs, we could determine the plausibility of Farmer’s evolutionary scenario. Unfortunately, we lack the data needed to quantify the beneﬁts and costs of endothermy for parental care (Angilletta and Sears 2003). Plus, Farmer’s hypothesis does not get around the problem associated with the classic thermoregulatory hypothesis: ultimately, either size or insulation would need to evolve for the transition to homeothermy. Alternatively, the initial selection of metabolic heat production might have had little (if anything) to do with thermoregulation. Instead, the initial target of selection might have been a greater maximal or sustained capacity for energy expenditure during activity, growth, or reproduction (Bennett and Ruben 1979; Koteja 2000). Proponents of this idea argued that selection of greater metabolic capacity would have caused a correlated increase in resting metabolic rate. According to Else and colleagues (2004), metabolic capacity and resting rates could be connected by the function of membranes, which appear to control rates of energy expenditure in cells. In this scenario, the heat produced by the ﬁrst endotherms would have played an insigniﬁcant role in thermoregulation. Later, adaptation to conserve heat would have facilitated the transition to endothermic homeothermy. Hypotheses based on an indirect response to selection rely entirely on the assumption that either pleiotropy or linkage exists between resting and maximal (or sustained) metabolic rates. Although these rates are often correlated (reviewed by Boily 2002; Hayes and Garland 1995), correlations can vary from strongly positive to strongly negative among closely related species (Gomes et al. 2004). We must also view these results cautiously because a phenotypic correlation does not necessarily possess the same magnitude or even the same sign as a genetic correlation (Hayes and Garland 1995). Recently, genetic correlations between resting and

121

active metabolic rates have been quantiﬁed for certain mammals, either through breeding experiments or artiﬁcial selection. Sadowska and colleagues (2005) analyzed a multigenerational pedigree of bank voles to estimate the genetic correlation between metabolic rates while resting and swimming. A strong positive genetic correlation was observed, supporting the assumption of the aerobic capacity model. Ksiazek and colleagues (2004) artiﬁcially selected mice for either low or high basal metabolic rate. After 18 generations of selection, they had caused a divergence in basal metabolic rate of 2.3 standard deviations between selection lines. Contrary to the prediction of Bennett and Ruben’s (1979) model, the maximal metabolic rate during swimming decreased as a correlated response to selection for a higher basal metabolic rate. Nevertheless, selection for high basal metabolic rate caused the evolution of greater food consumption and heavier internal organs; these results suggest that selection for high rates of sustained metabolism would have led to a correlated increase in basal metabolic rate, as suggested by Koteja (2000). Evaluating a scenario for the evolution of endothermic thermoregulation requires three elements. First, we must ask whether metabolic heat could have caused the hypothetical increase in body temperature in the hypothetical environment. For a well-deﬁned hypothesis, we could use mathematical models of heat transfer to answer this question, especially when we consider an ancestral organism unlike any living form (e.g., Phillips and Heath 2001). Second, we must argue convincingly that the beneﬁt of thermogenesis would have outweighed the cost. Finally, we must consider whether the net beneﬁt of endothermic thermoregulation would have exceeded the net beneﬁt of ectothermic thermoregulation. Unfortunately, the lack of mathematical models describing the evolution of endothermy prevents a rigorous test of existing ideas. Recall from Chapter 1 that verbal models come with implicit assumptions and usually make imprecise predictions. For example, the idea that endothermy evolved as an indirect response to direct selection for aerobic capacity has generated a ﬁxation on correlations between basal and maximal metabolic rates. But we know that genetic correlations change over time (Prasad and Shakarad

T H E R M A L A D A P TAT I O N

4.8.2 Optimal thermoregulation by endotherms Perfect homeotherms exist only in uniform environments, such as the depths of the ocean, and endothermic homeotherms are no exception. Endotherms generally exhibit bouts of hypothermia, ranging from a mild reduction in temperature during sleep to a drastic reduction in temperature during hibernation (Fig. 4.28). These hypothermic responses, commonly referred to as torpor, fall along a continuum; for example, the minimal temperature during hypothermia ranges from 4◦ C to 38◦ C among species of birds (McKechnie and Lovegrove 2002). Grigg and colleagues (2004) concluded that all degrees of torpor rely on similar physiological mechanisms related to ancestral features of physiology. Biologists agree that torpor was probably present in early mammals, but disagree on whether torpor was present in the ﬁrst birds or evolved repeatedly within this group (Geiser 1998; Grigg 2004; McKechnie and Lovegrove 2002). If torpor was an ancestral property of mammals and birds, we could view endothermic homeothermy as a facultative strategy adopted by organisms when the energetic beneﬁt

(a) 35

Temperature (°C)

34 33 32 31 30 29 28 27 26 25 7/4

7/6

7/8

7/10

7/12

7/14

(b) 40 35 30 25 20 15

2

1 2/

/3

/3

1/

12

11

4

5

/4 10

9/

6

7

6

8/

7/

6/

5/

7

10 4/

2004), such that any correlation in the ancestor of mammals or birds will not necessarily mirror the correlations existing today. How strong a correlation was necessary and for how long must it have persisted for endothermy to evolve? Once the incipient steps toward endothermy were taken, other developmental and genetic constraints might have kept organisms evolving towards homeothermic endothermy, even if genetic correlations between basal and maximal metabolic rates were lost in some lineages. Furthermore, direct selection for endothermy to expand the niche or care for offspring could have followed indirect selection for endothermy. Indeed, current hypotheses are not competing hypotheses in the sense that they mutually exclude one another. For this reason, biologists cannot conduct a critical experiment to distinguish among these hypotheses. Quantitative models of evolution will be necessary to evaluate the relative roles of hypothetical selective pressures leading to the origin of endothermy.

Temperature (°C)

122

Figure 4.28 In echidnas, intermittent periods of hypothermia result from (a) daily torpor and (b) seasonal hibernation. Reproduced from Grigg et al. (2004, Physiological and Biochemical Zoology, University of Chicago Press). © 2004 by The University of Chicago.

outweighs the obvious cost. This view turns our original question on its head: instead of asking why endothermy evolved from ectothermy, we could ask why ectothermy remained an important strategy for organisms capable of endothermy. In doing so, we must consider how the respective costs and beneﬁts of endothermy drive diel or seasonal shifts between alternative strategies. Huey and Slatkin’s model of optimal thermoregulation applies equally well to ectotherms and endotherms. Like an ectotherm, an endotherm would experience a decrement in performance if its body temperature were to deviate from its thermal optimum. Furthermore, the cost of thermoregulation should be a function of environmental temperature (Ames 1980). In fact, the beneﬁt and cost curves for endotherms could share the same properties as those curves considered by Huey and Slatkin (see Fig. 4.11), except the implicit mechanism for thermoregulation would be physiological rather than behavioral. Despite this fact, biologists have only

T H E R M O R E G U L AT I O N

recently applied quantitative models of energetic costs and beneﬁts to understand the timing and intensity of endothermy (but see Heinrich 1974). In this section, we shall consider optimal patterns of daily and seasonal torpor. Do endothermic organisms maintain less precise temperatures when environmental conditions reduce the energetic beneﬁts or raise the energetic costs? Based on the model, we should expect torpor to be more common when food is scarce or the environment is cold. The degree of torpor varies among species from different environments in ways that support this view. Species of ground squirrels express different degrees of metabolic depression, ranging from daily torpor in hot deserts to extreme hibernation in the cold Arctic (reviewed by Grigg 2004). Experimental manipulation of ambient temperature causes mammals to increase the use of torpor as temperature decreases (Ellison and Skinner 1992; Soriano et al. 2002). Similarly, food deprivation generally decreases the body temperatures of mammals (Piccione et al. 2002); in two experiments involving Siberian hamsters, minimal body temperature was positively related to food intake (Dark et al. 1996; Stamper et al. 1998). Injections of metabolic inhibitors (Westman and Geiser 2004) or murine leptin (Geiser et al. 1998) have experimentally linked the initiation of daily torpor to a lack of cellular energy. Birds, bees, and ﬂowers (but not trees) also switch from endothermy to ectothermy when the net beneﬁts of thermoregulation decline. Some birds enter torpor during periods of cold or starvation (reviewed by McKechnie and Lovegrove 2002). Bumblebees will use endothermy to sustain ﬂight only when the environment contains sufﬁcient nectar (reviewed by Heinrich 1974). Flowers of the skunk cabbage abandon endothermic homeothermy when air temperature drops below a certain level (Seymour and Blaylock 1999). Based on factorial experiments, both cold and starvation combine to cause a greater reduction in body temperature than does either factor alone (Westman and Geiser 2004). Energetic status also dictates the timing and degree of seasonal torpor (or hibernation). Because hibernation involves much greater depressions of metabolism and temperature, costs of survival stem not only from predators but also from physiological

123

damage to tissues and organs. The risk of physiological damage must be weighed against the risk of starvation. Humphries and colleagues (2003a) argued that hibernators would avoid deep and prolonged torpor if energetic resources were plentiful. Based on their reasoning, the intensity and duration of seasonal torpor should relate to the availability of food or the energy required for thermoregulation. Humphries and colleagues supported their idea with experimental evidence. Chipmunks that were supplementally fed prior to hibernation maintained higher temperatures during torpor and remained aroused for longer periods than did chipmunks belonging to an unfed control group (Humphries et al. 2003b). Furthermore, fat males tend to emerge from hibernation earlier than do skinny males and reproductive females (Dark 2005). Greater energy stores are less useful when they contribute indirectly to physiological damage. For example, an abundance of polyunsaturated fatty acids in membranes and fat bodies can lead to oxidative damage. The indirect cost associated with this energy source can offset the direct beneﬁt. In support of this view, chipmunks fed a diet poor in polyunsaturated fatty acids underwent longer and deeper bouts of torpor than did chipmunks fed a diet rich in polyunsaturated fatty acids (Munro et al. 2005). In addition to manipulations of environmental temperature and food intake, researchers have directly manipulated the energetic cost of hibernation. Kauffman and colleagues (2004) shaved ground squirrels 4 weeks into their hibernation and, as a control, similarly handled but did not shave other squirrels. Both shaved and control squirrels underwent multiday periods of torpor separated by homeothermic arousals lasting less than 24 hours. Shaved squirrels took longer to rewarm than control squirrels, presumably because of the greater loss of heat through their body surfaces. Surprisingly, shaved and control squirrels reached similar minimal body temperatures and remained in torpor or arousal for similar durations. Yet, the energetic cost of arousal did limit the duration of hibernation. Despite the fact that shaved squirrels consumed more food than control squirrels consumed during each arousal, shaved squirrels lost mass more rapidly and ceased hibernation several weeks earlier than control squirrels. Thus, the cost of rewarming

T H E R M A L A D A P TAT I O N

forced shaved squirrels to abandon seasonal torpor earlier in search of food to replenish their energy stores. Social interactions reduce the cost of daily or seasonal torpor, resulting in adaptive plasticity of endothermic thermoregulation. Huddling improves heat balance by minimizing convective heat loss and promoting radiative heat gain. Thus, a social animal would need to spend less energy to maintain a given body temperature than would a solitary animal. This observation leads to the prediction that social individuals should thermoregulate more precisely than solitary individuals. Consistent with this prediction, speckled mousebirds adjusted the depth of daily torpor according to the presence of conspeciﬁcs, as well as the abundance of food (McKechnie et al. 2006). Given surplus rations, birds in a communal roost maintained stable temperatures during the night, while birds roosting alone experienced steady declines in body temperature. When rations were restricted, both social and solitary birds cooled progressively during the night; however, solitary birds reached lower body temperatures than social birds. Plasticity of seasonal torpor also depends on social behavior. For example, skunks that wintered in isolation underwent more frequent and more intense bouts of torpor than did skunks that wintered in groups (Ten Hwang et al. 2007). Despite the difference in torpor, social skunks emerged from hibernation with more than twice the body fat possessed by solitary skunks. As with behavioral thermoregulation, endothermic thermoregulation depends on the net beneﬁts of nonenergetic currencies as well as energetic ones. Pravosudov and Lucas (2000) used stochastic dynamic programming to model the optimal degree of daily torpor when the energetic cost of endothermy trades off with its survival beneﬁt. They reasoned that birds in torpor were unusually susceptible to predation. Therefore, an individual must balance the risk of running out of energy during thermoregulation with the risk of falling prey during torpor. Although additional feeding could provide the energy needed to avoid torpor, this behavior would impose a risk of predation. Based on their model, the interaction among environmental temperature, predation risk, and foraging success strongly affects the optimal strategy of torpor (Fig. 4.29). Given little

(a) 1.0 0.8 0.6 Proportion of hypothermic birds

124

0.4 0.2 0

–30

–25

–20

–15

–10

–5

–30

–25 –20 –15 –10 Air temperature (°C)

–5

(b) 1.0 0.8 0.6 0.4 0.2 0

Figure 4.29 The interaction among environmental temperature, predation risk, and foraging success imposes a strong effect on the optimal strategy of torpor. These plots show the predictions of a stochastic dynamic model of optimal torpor in birds. (a) The optimal proportion of hypothermic birds at a particular air temperature depends on the risk of nocturnal predation (stars > triangles > diamonds > squares > circles). (b) The optimal proportion of hypothermic birds at a particular air temperature depends on the variance of foraging success (triangles > diamonds > squares > circles). Adapted from Pravosudov and Lucas (2000) with permission from Blackwell Publishing.

risk of predation, a bird should undergo daily torpor at any temperature. In a riskier environment, some torpor should occur at very low temperatures but no torpor should occur at higher temperatures. This result highlights a beautiful symmetry between models of ectothermic and endothermic thermoregulation: while ectotherms should forgo thermoregulation and its energetic beneﬁts to avoid predation (Section 4.7.2), endotherms should adopt thermoregulation and its energetic costs to avoid predation. Nevertheless, the optimal strategy of torpor depends on energy availability; when foraging success varies greatly, the optimal strategy likely involves some torpor during cold nights in spite of predation risk. Unlike the daily torpor of birds, the

T H E R M O R E G U L AT I O N

seasonal torpor of mammals might reduce the risk of predation. Hibernation often begins before the environment restricts energy acquisition, suggesting that predation risk causes animals to abandon homeothermy before energetic costs do. In fact, echidnas hibernate even in mild climates where food occurs year round. Grigg and Beard (2000) argued this behavior enables echidnas to avoid predators before breeding, once sufﬁcient energy reserves have accumulated. Their interpretation seems compelling because additional growth could occur if individuals were to delay hibernation. In light of these examples, we need more research to clarify the potentially contrasting roles of predation risk in the evolution of daily and seasonal torpor. Based on the ubiquity of torpor among endotherms, Grigg and Beard (2000) hypothesized that the ﬁrst endotherms were facultative homeotherms. These researchers also argued that we can learn much about the origin of endothermy by reﬂecting on the similarity and diversity of thermoregulatory strategies adopted by extant endotherms. Perhaps mathematical models tailored toward the costs and beneﬁts of endothermic thermoregulation will ultimately help to answer questions about both the maintenance and the origin of endothermy.

4.9 Conclusions Our evolutionary understanding of thermoregulation relies heavily on optimality models, ranging from Huey and Slatkin’s classic model to the more

125

recent models tailored to speciﬁc taxa. These models have guided several decades of research on the costs and beneﬁts of behavioral thermoregulation, and have recently been applied to endothermic thermoregulation. The theory has predicted the qualitative behaviors of animals extremely well when the costs of energy expenditure and mortality risk were manipulated in controlled environments. The theory also seems to explain the endothermic thermoregulation of birds and mammals in experiments conducted in artiﬁcial and natural environments. Overall, optimality models have been more successful at predicting thermoregulation than they have been at predicting thermal sensitivity (see Chapter 3). Perhaps for this reason, we know far less about the genetic constraints on thermoregulatory strategies than we do about genetic constraints on performance curves. Despite the success of optimality models when applied to experimental systems, our understanding of behavioral thermoregulation in natural environments suffers from a failure to characterize thermal constraints and an inability to estimate diverse costs. Recent advances in mapping operative environments (Chapter 2) and simulating thermoregulatory costs (see Box. 4.4) should take us beyond the insights gained from experiments in artiﬁcial environments. These approaches will also enable us to consider the interactions among the costs of energy expenditure, mortality risk, and missed opportunities. Thus, we are poised to build on the progress of the past.

CHAPTER 5

Thermal Acclimation

All organisms possess some capacity to modify their behavioral, physiological, or morphological characteristics in response to environmental temperature. Brief exposure to extreme heat or cold often causes greater tolerance of thermal extremes within hours (Hoffmann et al. 2003; Sinclair and Roberts 2005; Sinclair et al. 2003), a phenomenon called hardening. Prolonged exposures to moderate temperatures can trigger lasting changes in thermoregulation or thermosensitivity (Fig. 5.1). In this way, gradual shifts in the mean and variance of environmental temperature among seasons lead to responses that enhance performance (e.g., Hazel 2002; Packard et al. 2001; Sinclair 1997). Generally, both rapid and gradual responses to environmental temperature are reversible, but some responses remain ﬁxed throughout the life of the organism (Johnston and Wilson 2006). Collectively, I refer to these examples of phenotypic plasticity as thermal acclimation.11

11

Other researchers have deﬁned thermal acclimation differently (Bowler 2005; Lagerspetz 2006; Wilson and Franklin 2002). Two features distinguish competing deﬁnitions. First, some researchers regard the period of thermal exposure (i.e., the treatment) as acclimation, whereas others regard the change in phenotype (i.e., the response) as acclimation. Second, some researchers who deﬁne acclimation as a change in phenotype also require that this response improves the ﬁtness of the organism. I favor an intermediate position. In this book, I use thermal acclimation to refer to any phenotypic response to environmental temperature that alters performance and plausibly enhances ﬁtness. Therefore, I do not require precise knowledge of the ﬁtness consequences because such data rarely exist. Finally, I do not distinguish between responses to acute and chronic exposures to thermal extremes, and thus consider heat and cold hardening as forms of acclimation (for further discussion of this topic, see Loeschcke and Sorenson 2005).

126

To be precise, I distinguish between two types of thermal acclimation: developmental acclimation and reversible acclimation. Developmental acclimation comprises irreversible responses to temperatures experienced during ontogeny, whereas reversible acclimation comprises regulated responses to diel or seasonal changes in temperature. As we shall see, this distinction becomes critical when discussing the evolution of acclimation. Unfortunately, we cannot always distinguish between these forms of acclimation in practice. First, similar mechanisms could underlie the detection of and response to thermal heterogeneity in both

2.5

Filtration rate (L g–1 h–1)

5.1 Patterns of thermal acclimation

2.0

1.5

1.0

5°C 10°C

0.5

15°C 1

7

14 Day

21

28

Figure 5.1 Acute changes in temperature caused predictable changes in the feeding performance of mussels (Mytilus edulis), but prolonged exposures to novel temperatures triggered physiological responses that gradually restored performance to its original level. Data are mean ﬁltration rates of individuals exposed to 5◦ C, 10◦ C, or 15◦ C. Error bars are standard deviations. Adapted from Widdows and Bayne (1971) with permission from Cambridge University Press.

T H E R M A L A C C L I M AT I O N

processes, even though acclimation during early and late stages of ontogeny can differ dramatically (Johnsen et al. 2005; Terblanche and Chown 2006; Zeilstra and Fischer 2005). Second, we never know whether acclimation is truly irreversible; we only know that the time required for reversal exceeds the lifespan or that the alleles required to regulate reversal have not arisen. Researchers have compounded this problem by failing to design experiments that separate reversible acclimation from developmental acclimation (Wilson and Franklin 2002). For these reasons, some choose to disregard any categorization of acclimation responses. Indeed, we could place all patterns of acclimation along a continuum, deﬁned by the rate of the response or the potential to reverse the response; for example, diel and seasonal changes bound a continuum deﬁned by the rate of the response (Loeschcke and Sorenson 2005). At the very least, all acclimation responses share some common properties: the detection of environmental signals, the transduction of this signal into a cellular response, and the activation of the molecules (e.g., genes, polymerases, ribosomes, enzymes) that cause a change in phenotype (Angilletta et al. 2006a; Wilson and Franklin 2002). In this chapter, we shall examine the acclimation of performance curves as an adaptation to thermal heterogeneity. In doing so, we must confront some challenging questions. Does acclimation to high temperatures impair performance at low temperatures, and vice versa? If not, what other costs do organisms incur during acclimation? Do most acclimation responses enhance the ﬁtness of organisms or are these responses merely unavoidable consequences of thermal change? What kinds of thermal environments favor acclimation and what kinds favor alternative strategies? To address these questions, we need to consider the beneﬁts and costs of acclimation in variable and unpredictable environments. Let us start by considering the beneﬁts of acclimation.

5.2 The beneﬁcial acclimation hypothesis Although biologists have documented a broad diversity of acclimation responses in ectotherms and endotherms (Chown and Nicolson 2004; Hutchison

127

and Maness 1979; McNab 2002), the ﬁtness beneﬁts of these responses have generally been assumed rather than demonstrated (Huey and Berrigan 1996). Imagine an organism that encounters a gradual decline in environmental temperature. This individual could decrease its thermal optimum for performance or modify structures needed to enhance thermoregulation (e.g., thicken its pelage or darken its surface). These changes seem obviously beneﬁcial. Yet not all responses to thermal change enhance the performance of an organism. For example, the term deacclimation commonly refers to the reversal of a previous acclimation response. Because deacclimation often occurs rapidly (Kalberer et al. 2006), this response can be maladaptive when the environment ﬂuctuates wildly. In the last decade, biologists have made a much greater effort to establish whether acclimation responses actually beneﬁt organisms. The assumption that all acclimation responses enhance ﬁtness has come to be known as the beneﬁcial acclimation hypothesis. According to this hypothesis, a change in the environment of an organism should elicit a change in the phenotype that enhances performance in the new environment (Leroi et al. 1994). The beneﬁcial acclimation hypothesis has been tested many times, generally following a standard protocol. First, organisms are held under controlled environmental conditions for an initial period. Then the same individuals are shifted to one or more new environmental conditions for some period. Finally, the performance of each individual is assessed over a range of environmental conditions to determine whether performance in the new environment has been enhanced. Despite its intuitive appeal, the beneﬁcial acclimation hypothesis makes predictions that often deviate from the results of experiments. To substantiate this claim, I must brieﬂy review some studies of developmental and reversible acclimation.

5.2.1 Developmental acclimation Two forms of developmental acclimation could beneﬁt organisms in thermally heterogeneous environments. First, organisms could respond to temperatures during one stage of the life cycle to enhance performance during a subsequent stage (intragenerational response). Second, parents could respond

128

T H E R M A L A D A P TAT I O N

to temperatures during their life to enhance the performance of their offspring (intergenerational response); this response requires a parent to supply its offspring with nutrients, hormones, mRNA, or other factors that alter the physiological state at birth. Both forms of developmental acclimation have been the focus of recent attempts to evaluate the beneﬁcial acclimation hypothesis. Plant physiologists have a long history of investigating developmental acclimation of photosynthetic rate, focusing primarily on intragenerational responses (reviewed by Berry and Bjorkman 1980). In most studies, researchers raised individuals at constant temperatures or constant diel cycles for some portion of development (e.g., from the germination stage to the seedling stage). After this period, net photosynthetic rates were measured over a range of temperatures that included the developmental temperatures. Because physiologists report photosynthetic rates for a given area or mass of leaves, differences in these rates between plants from different developmental environments reﬂect acclimation of physiology rather than plasticity of biomass (although the allocation of biomass also contributes to photosynthetic acclimation; Atkin et al. 2006a). Collectively, these experiments provide a wealth of evidence that pertains to the beneﬁcial acclimation hypothesis. And most have generated patterns that differ markedly from those predicted by this hypothesis (Table 5.1). Quite often, plants that developed at a particular temperature performed best across a wide range of temperatures (Fig. 5.2). Interestingly, the diversity of performance curves produced within populations by developmental acclimation resembles that which has been produced among populations by genetic divergence (see Chapter 3). Although fewer animal physiologists have studied the developmental acclimation of performance curves, their research further conﬁrms the shortcomings of the beneﬁcial acclimation hypothesis. Zamudio and colleagues (1995) compared the social dominance of fruit ﬂies raised at 18◦ C and those raised at 25◦ C. Regardless of the temperature of their social environment, males raised at the higher temperature dominated males raised at the lower temperature. Rhen and Lang (1999) incubated chelonian embryos at 24◦ C, 26.5◦ C or 29◦ C and measured growth at 19◦ C and 28◦ C after hatching. At both

temperatures, hatchlings from the colder environments grew faster than hatchlings from the warmest environment. Both studies revealed optimal temperatures for development rather than the beneﬁcial acclimation of performance. Escherichia coli and Drosophila melanogaster have been useful models for studies of intergenerational responses. In a highly inﬂuential experiment, Leroi and colleagues (1994) grew lines of E. coli at either 32◦ C or 41.5◦ C for 24 hours and then competed cells from these lines at the same temperatures. Regardless of the competitive environment, the descendents of bacteria grown at 32◦ C outcompeted the descendents of bacteria grown at 42◦ C. Nevertheless, the descendents of bacteria grown at 42◦ C survived longer when exposed to the stressful temperature of 50◦ C. Crill and colleagues (1996) used a factorial manipulation of developmental temperatures experienced by parents and offspring of D. melanogaster. They crossed males and females that were raised at either 18◦ C or 25◦ C and then raised some offspring at each temperature. The knockdown temperatures of offspring were unaffected by the father’s developmental temperature and were only weakly affected by the mother’s developmental temperature. However, offspring raised at 25◦ C tolerated higher temperatures than did offspring raised at 18◦ C. Similarly, the thermal sensitivity of walking speed was affected by the developmental temperature but not by parental temperature; ﬂies raised at 25◦ C had a lower thermal optimum and a wider performance breadth than ﬂies raised at 18◦ C. In contrast, a similar experiment revealed that the ﬁtness of D. melanogaster was affected primarily by the developmental temperature of parents, with parents from warmer environments giving rise to offspring with greater ﬁtness across a wide range of temperatures (Gilchrist and Huey 2001). Despite these inconsistencies, intergenerational studies also suggest an optimal temperature for development. Based on the majority of evidence gathered to date, the beneﬁcial acclimation hypothesis offers few insights about the physiological responses to developmental temperature. Instead, an optimal temperature for development seems to exist, at which organisms that develop at a certain temperature outperform those that develop at other

T H E R M A L A C C L I M AT I O N

129

Table 5.1 Representative studies of the thermal acclimation of photosynthetic rate (adjusted for the area or mass of leaves). For each study, I note whether the results support the beneﬁcial acclimation hypothesis (BAH). Acclimation temperatures

Test temperatures

Beneﬁcial acclimation?

1–2◦ C above ambient temperature vs. ambient temperature (control) 1–2◦ C above ambient temperature vs. ambient temperature (control) 0, 5

A range of 14◦ C centered near the ambient temperature A range of 14◦ C centered near the ambient temperature 0, 5, 10, 15, 20, 25

No

Zhou et al. (2007)

No

Zhou et al. (2007)

Cenchrus ciliaris

25/20, 35/30

5, 10, 15, 20, 25, 30, 35, 40

No

Colobanthus quitensis

7/7, 12/7, 20/7

2–30◦ C in various intervals

No

Deschampsia antarctica

7/7, 12/7, 20/7

2–35◦ C in various intervals

No

Erigeron candensis

5, 25

Yes

Flaveria bidentis

25/20, 35/30

5–50◦ C and 20–50◦ C in cold-acclimated and warm-acclimated plants, respectively 5, 10, 15, 20, 25, 30, 35, 40

Gymnogongrus antarcticus

0, 5, 10

0, 5, 10, 15, 20, 25

No

Helianthis mollis

1–2◦ C above ambient temperature vs. ambient temperature (control)

A range of 14◦ C centered near the ambient temperature

No

Kallymenia antarctica

0, 5

0, 5, 10, 15, 20, 25

No

Panicum coloratum

25/20, 35/30

25, 28, 31, 34, 37, 40

No

Species Ambrosia psilostachya Aster ericoides Ballia callitricha

No

No

Comments

Algae grown at 0◦ C performed better at all temperatures. Plants grown at low temperatures outperformed plants grown at high temperatures. Plants grown at 12◦ C/7◦ C performed best over the range of 2–35◦ C. Plants exposed to 12◦ C performed best at all temperatures.

Source

Eggert and Wiencke (2000) Dwyer et al. (2007)

Xiong et al. (2000) Xiong et al. (2000) Dulai et al. (1998)

Plants grown at low temperatures outperformed plants grown at high temperatures. Algae grown at 0◦ C performed as well as or better than algae grow at 5◦ C or 10◦ C. In summer, Topt was greater for warmed plants than for control plants.

Plants grown at low temperatures outperformed plants grown at high temperatures.

Dwyer et al. (2007)

Eggert and Wiencke (2000) Zhou et al. (2007)

Eggert and Wiencke (2000) Dwyer et al. (2007)

Continued

130

T H E R M A L A D A P TAT I O N

Table 5.1 Continued

Species

Acclimation temperatures

Test temperatures

Beneﬁcial acclimation?

Pisum sativum

25/18, 35/18

25, 30, 35, 40, 45

No

Phyllophora ahnfeltioides

0, 5, 10

0, 5, 10, 15, 20, 25

No

Ruppia drepanensis

10, 20

10, 20, 30

Partial

Sorghastrum nutans

1–2◦ C above ambient temperature vs. ambient temperature (control)

A range of 14◦ C centered near the ambient temperature

No

Spinacia oleracea Triticum aestivum

30/25, 15/10

10–40◦ C in various intervals

Yes

15, 25, 35

5, 10, 15, 20, 25, 30, 35, 40

Partial

Valonia utricularis

15, 18, 20, 25

5, 10, 15, 20, 25, 30, 35

No

Xylia xylocarpa

25/20, 30/20, 35/20, 40/20

20, 25, 30, 35, 40, 45, 50

No

temperatures (Huey and Berrigan 1996; Huey et al. 1999). Some researchers believe this trend results from the artiﬁcial conditions used in acclimation experiments; in other words, when an organism must undergo development at a constant temperature, the detrimental effect of development at an extreme temperature probably overwhelms any effect of thermal acclimation (Wilson and Franklin 2002). This argument gains some support when we contrast experiments conducted at constant temperatures with a recent experiment conducted

Comments

When measured at 30◦ C, algae grown at 30◦ C outperformed algae grown at 20◦ C. In summer, Topt was greater for warmed plants than for control plants.

Plants grown at 15◦ C and 25◦ C outperformed others at their respective developmental temperatures; however, leaves grown at 35◦ C performed poorly at all temperatures. Algae grown at 18◦ C outperformed algae grown at all other temperatures. Seedlings grown at 40◦ C/20◦ C outperformed other seedlings at 0◦ C and 50◦ C.

Source Haldimann and Feller (2005) Eggert and Wiencke (2000) Santamaría and Hootsmans (1998) Zhou et al. (2007)

Yamori et al. (2006) Yamasaki et al. (2002)

Eggert et al. (2003b)

Saelim and Zwiazek (2000)

at ﬂuctuating temperatures. Bilcke and colleagues (2006) incubated lizard eggs at either warm or cool diel cycles and compared the feeding and digestive performances of hatchlings. For most measures of feeding performance, lizards that developed in the cooler environment performed better at 20◦ C, but the reverse was true at 28◦ C. Bear in mind that these studies of developmental acclimation probably confounded irreversible and reversible responses to temperature. In each experiment, we do not know whether the organisms

T H E R M A L A C C L I M AT I O N

5

(a) A. glabriuscula

(b) A. sabulosa

(c) T. oblongifolia

(d) A. lentiformis

(e) N. oleander

(f) L. divaricata

131

4

Photosynthetic rate (nmol cm–2 s–1)

3 2 1 0 4 3 2 1 0

0

20

30

40

50 10 20 30 40 50 10 Leaf temperature (°C)

20

30

40

50

Figure 5.2 The developmental acclimation of net photosynthetic rate varied greatly among six species of plants (Atriplex glabriuscula, Atriplex sabulosa, Tidestromia oblongifolia, Atriplex lentiformis, Nerium oleander, and Larrea divaricata). Each panel shows performance curves of plants grown under cool and hot conditions (ﬁlled circles and open circles, respectively). Temperatures during development are indicated by the vertical dashed lines. Data in (d–f) support the beneﬁcial acclimation hypothesis, but those in (a–c) do not. Reprinted, with permission, from the Annual Review of Plant Physiology (Berry and Bjorkman 1980), Volume 31. © 1980 by Annual Reviews (www.annualreviews.org).

could have reversed their responses to developmental temperature. By investigating the cellular mechanisms of change, one might ﬁnd clues about the reversibility of these acclimation responses. Alternatively, one could simply attempt to reverse the effects of acclimation, but such attempts should be conﬁned to a single ontogenetic stage to ensure comparable measures of performance during acclimation and deacclimation.

5.2.2 Reversible acclimation The thermal acclimation of mature organisms has been a focus of physiologists for many decades. Presumably, changes in the capacity for performance during adulthood can be reversed if given enough time. Although this assumption has rarely been

tested, I suspect most acclimation responses during adulthood reﬂect processes that occur during diel or seasonal change. By comparing the performance curves of adults exposed to different temperatures, we can infer whether acclimation enhances performance as described by the beneﬁcial acclimation hypothesis. The wealth of evidence clearly shows that acclimation enhances some performances in some species (Huey et al. 1999). Of the experiments summarized in Table 5.2, almost half revealed clear beneﬁts of acclimation across a range of temperatures. Additional experiments revealed beneﬁts of acclimation at some temperatures but not others. Nevertheless, acclimation was rarely perfect in the sense that thermal optima corresponded to environmental temperatures. Only in one exceptional case was perfect acclimation observed; after 28 days of exposure

132

T H E R M A L A D A P TAT I O N

Table 5.2 Representative studies of thermal acclimation in ectothermic animals. For each study, I note whether the results support the beneﬁcial acclimation hypothesis (BAH).

Species

Performance

Acclimation temperatures

Test temperatures

Beneﬁcial acclimation?

Mollusks Ostrea edulis

Filtration

5, 10, 15, 20, 25

5, 10, 15, 20, 25, 30

No

Walking

10, 25

10, 25

Yes

Fighting success and chela force

20, 30

20, 30

Yes/No

Swimming and chela force

12, 20, 30

12, 20, 30

No

Swimming

7, 25

7,13, 19, 25

Partial

Carassius auratus

Swimming

10, 35

5, 10, 15, 25, 30, 35, 40

Yes

Carassius auratus

Swimming

5, 15, 25, 35

5, 10, 15, 20, 25, 30, 35, 38

Yes

Cyprinus carpio

Swimming

8, 26

10, 20

Partial

Fundulus heteroclitus

Swimming

10, 35

5, 10, 15, 25, 30, 35

Yes

Gambusia holbrooki Gambusia holbrooki

Mating success

18, 30

18, 30

Yes

Swimming

18, 30

18, 30

Yes/Partial

Gasterosteus aculeatus

Swimming

18, 23

18, 22

Partial

Tardigrades Macrobiotus harmsworthi Crustaceans Cherax destructor Cherax dispar

Fishes Barbus barbus

Comments

Source

Topt was related to acclimation temperature.

Newell et al. (1977)

Li and Wang (2005) Fighting success supported BAH, but chela force did not.

Seebacher and Wilson (2006) C. L. Bywater and R. S. Wilson, unpublished observations

Critical speed supported BAH, but voluntary and maximal speeds did not. Topt was related to acclimation temperature. Topt was related to acclimation temperature. Fish exposed to 8◦ C performed better at 10◦ C than ﬁsh exposed to 26◦ C, but both performed similarly at 20◦ C. Topt was related to acclimation temperature.

Low acclimation temperature had no effect in females. Fish exposed to 23◦ C swam faster at high temperatures than ﬁsh exposed to 18◦ C in one of two seasons.

O’Steen and Bennett (2003)

Johnson and Bennett (1995) Fry and Hart (1948) Rome et al. (1985)

Johnson and Bennett (1995) Wilson et al. (2007) Wilson et al. (2007) Guderley et al. (2001)

Continued

T H E R M A L A C C L I M AT I O N

133

Table 5.2 Continued Acclimation temperatures

Test temperatures

Beneﬁcial acclimation?

Swimming

5, 10, 15, 20, 25, 30, 35

5, 10, 15, 20, 25, 30, 35

Partial

Myoxocephalus scorpius

Swimming

5, 15

1, 5, 15, 20

Partial

Oncorhynchus kisutch

Swimming

2, 5, 8, 11, 14, 17, 20, 23

2, 5, 8, 11, 14, 17, 20, 23

Yes

Pagothenia borchgrevinki Pagothenia borchgrevincki Puntius schwanenfeldii

Swimming

–1, 4

–1, 2, 4, 6, 10

No

Swimming

–1, 4

–1, 2, 4, 6, 8

Yes

Swimming

17, 33

17, 20, 25, 30, 33

Partial

Taurulus bubalis

Swimming

5, 15

1, 5, 15, 20

No

Amphibians Bufo americanus

Jumping

5, 25

5, 10, 15, 20, 25

Partial

Swimming

10, 24

10, 24, 34

No

Species

Performance

Micropterus dolomieui

Limnodynastes peronii (tadpoles)

Comments

Source

Topt increased by 0.5◦ C per 1◦ C increment in acclimation temperature, over the range 5–25◦ C. Fish exposed to 15◦ C performed better at 15◦ C and 20◦ C than ﬁsh exposed to 5◦ C; however, treatments were conducted in different seasons. Topt increased by 0.25◦ C per 1◦ C increment in acclimation temperature.

Larimore and Duever (1968)

Critical speed supported BAH, but voluntary and maximal speeds did not. Fish exposed to 5◦ C performed better at 1◦ C than ﬁsh exposed to 15◦ C, but both performed similarly at 5◦ C and 15◦ C; however, treatments were conducted in different seasons. Toads exposed to 5◦ C jumped farther at 5◦ C than toads exposed to 25◦ C. At all temperatures, tadpoles exposed to 10◦ C swam as fast as or faster than tadpoles exposed to 24◦ C.

Temple and Johnston (1998)

Grifﬁths and Alderdice (1972)

Wilson et al. (2001) Seebacher et al. (2005) O’Steen and Bennett (2003)

Temple and Johnston (1998)

Renaud and Stevens (1983)

Wilson and Franklin (1999)

Continued

134

T H E R M A L A D A P TAT I O N

Table 5.2 Continued

Species

Performance

Acclimation temperatures

Test temperatures

Beneﬁcial acclimation?

Limnodynastes peronii (adults)

Jumping

18, 30

8, 12, 18, 23, 30, 35

No

Rana pipiens

Jumping

5, 25

5, 10, 15, 20, 25

Partial

Triturus dobrogicus

Swimming/running 15, 25

10, 15, 20, 25, 30, 33

Partial

Xenopus laevis (tadpoles) Xenopus laevis (adults)

Swimming

12, 30

5, 12, 20, 30

Yes

Swimming

10, 25

5, 10, 15, 25, 30

Partial

Frogs exposed to 10◦ C swam faster at low temperatures than frogs exposed to 25◦ C.

Reptiles Crocodylus porosus

Swimming

20, 29

17, 20, 25, 29, 32

Yes

Xantusia vigilis

Sprinting

20, 30

12.5, 15, 20, 25, 30, 34, 37.5

No

Topt matched the acclimation temperature. Acclimation beneﬁted performance at extreme temperatures.

Comments

Source

Frogs acclimated at 18◦ C would not jump at 35◦ C. Frogs exposed to 25◦ C jumped farther at 20◦ C and 25◦ C than frogs exposed to 5◦ C. Newts exposed to 25◦ C ran faster at high temperatures than newts exposed to 15◦ C.

Wilson and Franklin (2000b) Renaud and Stevens (1983)

Gvoždík et al. (2007)

Wilson et al. (2000b) Wilson et al. (2000b)

Glanville and Seebacher (2006) Kaufmann and Bennett (1989)

Notes: Topt , thermal optimum for performance.

to either 18◦ C or 29◦ C, crocodiles swam best at their respective acclimation temperature (Fig. 5.3a). More often, acclimation enhanced performance at one or more extreme temperatures without radically shifting the thermal optimum (Figs. 5.3b). And, in some cases, no beneﬁcial acclimation was observed (Fig. 5.3c). This diversity of ﬁndings calls for a more complex view than the beneﬁcial acclimation hypothesis. Thermal limits to performance acclimate more readily than thermal optima. For example, Kaufmann and Bennett (1989) exposed adult lizards to either 20◦ C or 30◦ C for 50 days and then compared the thermal sensitivity of sprint speed between groups (Fig. 5.4). The mean performance curves for these groups were very similar over the range of 15◦ C to 34◦ C, but diverged at extreme temperatures

(12.5◦ C and 37.5◦ C). Critical thermal limits had diverged in a manner consistent with the difference in sprinting performance. In fact, rapid and reversible acclimation of thermal tolerance occurs ubiquitously among animals. Acute exposure to extreme heat or cold results in greater tolerance upon subsequent exposure (see reviews by Chown and Nicolson 2004; Hoffmann et al. 2003; Sinclair and Roberts 2005). Interestingly, exposure to a moderate temperature can also alter the tolerance of extreme temperatures. In ﬁsh, the critical thermal limits scale linearly with water temperature over a range of 30◦ C (Fig. 5.5). Realistic diel cycles of temperature or light can induce rapid thermal hardening of fruit ﬂies (Kelty and Lee 2001; Sorenson and Loeschcke 2002). Acclimation of thermal limits without exposure to extreme temperatures seems

T H E R M A L A C C L I M AT I O N

1.2

Sustained speed (body lengths s–1)

(a)

counterintuitive because acclimation of the thermal optimum would be more beneﬁcial in moderate environments.

1.1 1.0 0.9

5.2.3 Beyond the beneﬁcial acclimation hypothesis

0.8 0.7 0.6 0.5

Swimming speed (m s–1)

16

(b)

18

20

22

24

26

28

30

32

1.6 1.2

*

*

*

0.8 0.4 0.0

(c)

0

5

5

10

10

15

20

25

30

35

15 20 25 30 Body temperature (°C)

35

1.2 Swimming speed (m s–1)

135

1.0 0.8 0.6 0.4 40

Figure 5.3 Acclimation responses of ectotherms range from perfect compensation to no compensation for a change in environmental temperature. (a) Estuarine crocodiles matched their thermal optimum for sustained swimming to their environmental temperature. Adapted from Glanville and Seebacher (2006) with permission from the Company of Biologists. (b) Clawed frogs altered their performance breadth for swimming speed in response to environmental temperature, but did not alter their thermal optimum. Adapted from Wilson et al. (2000) with kind permission from Springer Science and Business Media. © Springer-Verlag 2000. (c) Marsh frogs did not alter their thermal optimum or their performance breadth in response to environmental temperature Adapted from Wilson and Franklin (2000b) with permission from Elsevier. © 2000 Elsevier Ltd.

If we consider the magnitude and unpredictability of thermal heterogeneity in natural environments, the discord between the predictions of the beneﬁcial acclimation hypothesis and the observations of some researchers hardly seems surprising. The beneﬁcial acclimation hypothesis is a naïve, costfree view of nature. In other words, proponents of the beneﬁcial acclimation hypothesis have overlooked the fact that acclimation must impose some cost to an organism (Hoffmann 1995), just as other forms of phenotypic plasticity impose costs (DeWitt et al. 1998). In the remainder of this chapter, we shall abandon the beneﬁcial acclimation hypothesis in favor of mathematical models of optimal acclimation (Gabriel 1999; Gabriel and Lynch 1992). While the beneﬁcial acclimation hypothesis implicitly assumes that acclimation imposes no cost, alternative models explicitly consider costs and beneﬁts when predicting the optimal strategy of acclimation. Although optimality models of acclimation have been available for more than a decade, they have not been adopted by physiologists who study this phenomenon. Unfortunately, ignorance of these models has slowed the development of a quantitative theory of acclimation (see Huey and Berrigan 1996; Huey et al. 1999). By embracing these models, we can better understand why acclimation capacities have diverged among populations that experience different patterns of thermal heterogeneity.

5.3 Costs of thermal acclimation To model the optimal strategy of acclimation, one must have some knowledge of the costs of plasticity. In Chapter 3, I introduced three classes of tradeoffs that constrain performance curves: acquisition tradeoffs, allocation tradeoffs, and specialist– generalist tradeoffs (see Box 3.5). These same tradeoffs apply not only to genetic variation in performance curves among individuals but also to phenotypic variation within individuals. In other words, when an organism alters its performance

136

T H E R M A L A D A P TAT I O N

200 160 140 120

60 40 20 0

5

10

0

CTmax 20°C CTmax 30°C

80

CTmin 30°C

100 CTmin 20°C

Sprint speed (cm s–1)

180

15 20 25 30 35 Body temperature (°C)

40

Critical thermal limit (°C)

45 40 35 30 25 20 15 10 5 0

0

5

10 15 20 25 30 Acclimation temperature (°C)

35

40

Figure 5.5 The critical thermal limits acclimate readily during prolonged exposure to a moderate temperature. Each line represents the acclimation response for a different species of ﬁsh. Data were summarized by Beitinger and colleagues (2000).

curve via acclimation, one or more of these tradeoffs can reduce the net beneﬁt of this physiological response. Costs imposed by these tradeoffs follow from two facts. First, acclimation requires energy that an organism acquired at some risk and might otherwise use for a different function. Second, acclimation requires time for developmental or physiological changes to occur. Each of these requirements imposes costs in terms of survivorship or fecundity.

5.3.1 Costs of energetic demands Energetic costs arise from the detection of and response to thermal change. Detection initiates a series of molecular responses, ultimately leading to

45

50

Figure 5.4 The performance of lizards (Xantusia vigilis) at extreme temperatures responded to acclimation at either 20◦ C (ﬁlled symbols) or 30◦ C (open symbols). Adapted from Kaufmann and Bennett (1989, Physiological and Biochemical Zoology, University of Chicago Press). © 1989 by The University of Chicago.

modiﬁcations of cells and tissues. Cellular responses include the expression of new proteins (Schulte 2004; Somero and Hochachka 1971), the remodeling of cell membranes (Crockett 1998; Hazel 1995), and the operation of molecular chaperones (Feder and Hofmann 1999). From Chapter 3, recall that the functions of proteins and membranes deﬁne a performance curve. By changing these components of cells, an organism can alter its thermal sensitivity. For example, acclimation of photosynthetic capacity in plants involves the concentration of existing proteins and the expression of different isozymes (Yamasaki et al. 2002; Yamori et al. 2006). Similarly, acclimation of locomotor performance in animals involves changes in the quantities or qualities of metabolic enzymes (Rogers et al. 2004; Wilson and Franklin 1999). Changes in membrane ﬂuidity during thermal acclimation occur in both unicellular and multicellular organisms (Chintalapati et al. 2004; Lee et al. 2006; Overgaard et al. 2005). As with the genetic divergence of performance curves, greater membrane ﬂuidity enables better function at low temperatures (Fig. 5.6). Finally, acclimation to stressful temperatures usually involves antifreeze or heat-shock proteins (Lee and Costanzo 1998; Sorensen et al. 2003); for example, exposure to mild warming causes Antarctic ﬁshes to produce fewer glycoproteins (Jin and DeVries 2006) and Antarctic grasses to produce more Hsp70 (Reyes et al. 2003). Heat-shock proteins consume energy during their function, leading to costs beyond those required for production (Macario and de Macario 2007).

(a)

Arrhenius break temperature (°C)

T H E R M A L A C C L I M AT I O N

50

Green Pink Red Pinto

45 40 35 30

(b)

Fluorescence polarization at 20°C

0

0.34

5 10 15 20 25 Acclimation temperature (°C)

30

Green Pink Red Pinto

0.32 0.30 0.28 0.26 0.24 0

5 10 15 20 25 Acclimation temperature (°C)

30

Figure 5.6 Changes in the ﬂuidity of cellular membranes underlie the acclimation of physiological function. (a) For abalone of the genus Haliotis, acclimation alters the temperature at which mitochondrial respiration fails (the Arrhenius break temperature). (b) Acclimation also alters ﬂuidity of the mitochondrial membrane, as estimated by ﬂuorescence polarization. Abalone maintained at low temperatures possessed more ﬂuid membranes (and hence less polarization). Symbols represent different species of abalone. Error bars are 1 standard error of the mean. Reproduced from Somero (2002) with permission from Oxford University Press.

The energetic costs of these cellular responses are difﬁcult to quantify, but qualitative assessments suggest the costs are substantial (Somero 2002). Under certain conditions, these energetic costs will be manifested as specialist–generalist tradeoffs; for example, upregulation of new isozymes and downregulation of old ones would cause an increase in performance at some temperatures and a decrease in performance at others. Under other conditions, the energetic expense of acclimation would cause an acquisition or allocation tradeoff. In this way, acclimation to one range of temperature could occur

137

without loss of performance at other temperatures. In fact, an organism could even gain some capacity for performance at temperatures to which they have not been exposed (e.g., an increase in cold resistance during exposure to heat). In such cases, the energetic cost of acclimation would be manifested as a decrease in survivorship or fecundity. In summary, physiological acclimation involves some of the same constraints faced during evolutionary adaptation. Studies of phenotypic and genetic correlations associated with heat hardening have provided conﬂicting data about the costs of acclimation. In some experiments, heat hardening reduced fecundity (Krebs and Loeschcke 1994a) or longevity (Scott et al. 1997), but other experiments revealed no detrimental effects. For example, heat hardening of parasitic wasps (Trichogramma carverae) reduced the rate of parasitization in one study (Hercus et al. 2003), but enhanced parasitization without an obvious cost in another study (Thomson et al. 2001). This diversity of phenotypic correlations might reﬂect an underlying diversity of genetic correlations. Analyses of isofemale lines of Drosophila melanogaster revealed a negative relationship between heat hardening and survivorship (Krebs and Feder 1997), but artiﬁcial selection for heat hardening resulted in greater longevity (Norry and Loeschcke 2003). Moreover, isofemale lines of D. buzzatii exhibited a positive relationship between heat hardening and fecundity (Krebs and Loeschcke 1999). A recent study suggests these contrasting results might have stemmed from differences in the timing and intensity of heat stress. Repeated exposure of D. melanogaster to mild heat stress either enhanced or reduced its fecundity, depending on the age of the ﬁrst exposure (Hercus et al. 2003). Genetic engineering offers a more sophisticated approach to exploring potential costs of acclimation. Roberts and Feder (2000) engineered fruit ﬂies with extra functional copies of hsp70 and control ﬂies with nonfunctional copies. Flies with extra genes survived better than control ﬂies during moderate to severe heat stress, but the reverse was true in the absence of heat stress (Fig. 5.7). Other transgenic ﬂies with extra copies of hsp70 had lower fertility, primarily because their eggs hatched poorly (Silbermann and Tatar 2000). The poor survivorship and fecundity of transgenic ﬂies might have resulted

T H E R M A L A D A P TAT I O N

Temperature (°C)

(a)

36

27 Sept

32 28 10 Sept

24

(b)

45

Temperature (°C)

138

40 35 30

Field

25

Simulation of field

20

20 6

10

14

0

18

60

90

Time (min)

Time (CDT) (c)

30

100

Survivorship (%)

80

60

40

20

0

22

10

25°C constant in laboratory

22

10

Mild 10 Sept in field

22

10

Moderate 27 Sept in field

22

10

Severe Laboratory simulation of field

Thermal conditions Figure 5.7 The expression of heat-shock proteins by transgenic fruit ﬂies (Drosophila melanogaster) carries a cost under mild heat stress or no heat stress. (a) Temperatures experienced by larvae developing in rotting apples can rise above 32◦ C for varying durations. Temperatures are shown for days of mild and moderate heat stress (10 and 27 September, respectively; CDT, central daylight time).(b) Temperatures experienced during severe heat stress in apples were simulated in the laboratory. (c) Survivorship of transgenic ﬂies (shaded bars) exceeded that of control ﬂies (open bars) during moderate to severe heat stress, but the reverse was true in the absence of heat stress. Reproduced from Roberts and Feder (2000) with permission from Blackwell Publishing.

from energetic demands of heat-shock proteins or detrimental effects on development (Hoffmann et al. 2003). An elegant experiment by Krebs and Feder (1998) suggested that energetic costs of producing heat-shock proteins do not affect longevity. These researchers used lines of ﬂies genetically engineered to express β-galactosidase instead of hsp70. Repeated heat shock of larvae decreased

survivorship to adulthood, but the effect was greater at high rations than low rations; also, gene expression was unrelated to survivorship. Although one can interpret these ﬁndings as evidence that heat hardening costs little energy, energetic costs might have affected fecundity rather than survivorship. Furthermore, the function of heat-shock proteins could cost more than their production (Somero

T H E R M A L A C C L I M AT I O N

2002), and Krebs and Feder did not quantify this cost in their experiment. Despite the discrepancies among ﬁndings, acclimation clearly imposes costs of survivorship and fecundity in some cases. In other cases, perhaps a focus on animals raised under benign conditions makes tradeoffs less detectable. Although studies of heat hardening have given us an excellent start, we must devise creative ways to estimate the energetic cost of other acclimation responses. Undoubtedly, all forms of acclimation impose some energetic cost. The question for the future will be whether these costs outweigh others, such that acquisition and allocation tradeoffs greatly inﬂuence the evolution of acclimation. With this in mind, we now turn to a very different cost of acclimation: the cost arising from a time lag.

5.3.2 Costs of time lags While scientists have mainly studied acclimation to constant temperatures, most natural environments are anything but constant. Stochastic variations over space and time leave many organisms with great doubt about the future of their environment (see Chapter 2). If acclimation occurs rapidly, an organism could continuously adjust its physiology to deal with its current environment. But if acclimation occurs slowly, the environment will likely change again before the organism can adjust to current conditions. Therefore, the net beneﬁt of acclimation depends on the magnitude and predictability of diel, daily, and seasonal variations in temperature (see Kingsolver and Huey 1998). Obviously, the impact of thermal stochasticity on the cost of acclimation depends strongly on the time required for acclimation. We can infer the existence of time lags from repeated measures of performance during thermal change. For example, Widdows and Bayne (1971) showed that mussels required at least 28 days to restore their rate of ﬁltration after an abrupt change in temperature (see Fig. 5.1). In contrast, the acclimation of photosynthetic rate in some plants occurs within 2 weeks (Cunningham and Read 2003a). Tolerance of extreme temperatures can change throughout the course of a day, presumably because of the rapid cellular responses involving stress proteins (Buchner and Neuner 2003; Kelty

139

and Lee 2001; Sorenson and Loeschcke 2002). Even when thermal acclimation occurs rapidly, the precise rate differs among closely related species (see, for example, Podrabsky and Somero 2006). Although available data afford crude comparisons of time lags among species, we know virtually nothing about the extent to which time lags vary among genotypes within species. Given specialist–generalist tradeoffs, time lags will always impose some cost of acclimation. Specialization for the current temperature will lead to maladaptation when the environment changes more rapidly than the organism can respond. A generalist could avoid the problems associated with a time lag, but this strategy imposes other costs (see Chapter 3). The inﬂuence of time lags on the evolution of acclimation cannot be captured by simple ideas, such as the beneﬁcial acclimation hypothesis. Instead, we must adopt mathematical models developed speciﬁcally for the study of phenotypic plasticity in stochastic environments.

5.3.3 Interaction between costs In some cases, energetic costs and time lags depend on one another. As an example, consider the production of hsp70 following a heat shock. An organism can use one of two strategies to regulate the expression of heat-shock proteins: feedforward and feedback. Using feedforward regulation, the organism would produce a certain quantity of hsp70 immediately after heat shock. With this mechanism, new protein can function with minimal delay but uncertainty about the actual damage could lead to a wasteful overproduction of heat-shock proteins. Using feedback regulation, an organism would produce heat-shock proteins in direct response to denatured proteins. This mechanism ensures the production of hsp70 will match the need, but the organism suffers a lag in protection. Shudo and colleagues (2003) modeled the optimal combination of feedforward and feedback mechanisms of regulation. They assumed selection acts to minimize the sum of protein damage and production costs. Using this optimality criterion, an organism should always use feedforward regulation to some degree but should not necessarily use feedback regulation. Feedback regulation confers an

140

T H E R M A L A D A P TAT I O N

advantage only when the organism cannot accurately predict the degree of damage caused by heat shock. In fact, an organism should rely entirely on a feedforward mechanism when the degree of heat damage corresponds to the degree of heat exposure or when expression of hsp70 requires considerable time. Given poor predictably of damage and rapid upregulation of proteins, the relative degree of feedforward regulation versus feedback regulation depends on the actual damage caused by heat shock. Based on this model, the relative costs imposed by energetic demands and time lags should partly depend on the thermal environment, which determines the predictability and magnitude of heat stress.

5.4 Optimal acclimation of performance curves After reviewing the beneﬁts and costs of acclimation, we should understand the need to consider hypotheses about optimal acclimation. Obviously, a genotype that can assess its thermal environment and express the appropriate phenotype would perform better than a genotype that expressed a certain phenotype regardless of its environment. But the net beneﬁt of acclimation depends on the energy required for change and the predictability of the environment. What if acclimation requires considerable energy, thereby negatively affecting other traits that contribute to ﬁtness? And what if today’s environment correlates imperfectly with tomorrow’s environment? How do such constraints affect the optimal acclimation of performance curves? To answer these questions, we must turn to mathematical models plucked from the burgeoning literature on phenotypic plasticity. These models differ in their assumptions about environmental conditions and phenotypic strategies; therefore, let us brieﬂy consider these assumptions before diving into some examples. When modeling the evolution of phenotypic plasticity, theorists characterize environments as either ﬁne-grained or coarse-grained. In a ﬁne-grained environment, an individual experiences variation in environmental states during its life. In a coarsegrained environment, an individual experiences a single environmental state, but different individuals

experience different environmental states. Coarsegrained environments arise in one of two ways: (1) the environment varies temporally, but changes on a scale that exceeds the longevity of an individual, or (2) the environment varies spatially, but each individual settles within a homogeneous patch. In either kind of environment, some form of acclimation could enhance ﬁtness. Theorists characterize phenotypes as being irreversible or reversible. Irreversible phenotypes become ﬁxed early in ontogeny. The developmental environment provides cues about the future, including but not limited to temperature. These cues trigger the expression of a particular phenotype, which then remains constant thereafter. The cues provided by the developmental environment may or may not accurately reﬂect the temperatures of the selective environment. In contrast, reversible phenotypes can acclimate and deacclimate repeatedly. For real organisms, these phenotypic transitions occur gradually (see Fig. 5.1). For theoretical organisms, acclimation and deacclimation are more conveniently treated as abrupt transitions, subject to a time lag between phenotypic states. Obviously, reversible acclimation enables an organism to enhance its ﬁtness by continuously tuning its physiology to its environment. In the models considered here, the phenotypes consist of parameters that deﬁne a performance curve. Before considering any mathematics, our intuition should lead us to the following conclusions: (1) stable environments favor specialists without the capacity for acclimation; (2) environments that remain stable within generations but vary among generations favor developmental acclimation; and (3) environments that vary within generations favor reversible acclimation. Our intuition provides fewer insights about the optimal magnitude of acclimation, particularly for complex traits such as performance curves. First, any or all parameters of a performance curve might acclimate during ontogeny. Second, real environments vary spatially and temporally, and hence most organisms perceive their environment as both ﬁne-grained and course-grained. Given these complications, we must supplement our intuition with insights from mathematical models.

T H E R M A L A C C L I M AT I O N

5.4.1 Optimal developmental acclimation

(5.1)

where gi equals the genetic value of the thermal optimum or performance breadth, and ei represents developmental noise. To model acclimation, Gabriel and Lynch added a term (ri ) to eqn 5.1 that represents an irreversible response to the developmental environment: zi = gi + ei + ri .

(5.2)

If acclimation occurs without cost or constraint the optimal value of ri equals zi∗ − gi ,

(5.3)

where z∗i equals the optimal phenotype. However, Gabriel and Lynch assumed the capacity for acclimation was constrained. Speciﬁcally, they assumed that small changes in the performance curve were more feasible than large changes, and very large changes were impossible. This constraint was implemented by choosing a bounded, nonlinear reaction norm for ri (Fig. 5.9). Besides the addition of this acclimation response, the assumptions of this new model followed those of the earlier model (see Table 3.2); the most notable of these assumptions being that ﬁtness depends on the product of performance over time. In Gabriel and Lynch’s model, developmental acclimation of the thermal optimum increases the

g1

z1 shifted

Survivorship

0.4 0.3

g2

0.2 0.1 0.0 –2

0

2

4

(b) 0.5 0.4 Survivorship

Expanding on their previous work, Gabriel and Lynch (1992) modeled the optimal acclimation of a performance curve in an unpredictable environment. From Chapter 3, recall these researchers modeled the thermal sensitivity of survivorship with a Gaussian function (see Fig. 3.15). Two parameters—the thermal optimum (z1 ) and performance breadth (z2 )—deﬁned the speciﬁc shape of the curve (Fig. 5.8). Because the Gaussian function describes a probability distribution, the area under the curve remains constant for all values of z1 and z2 . This function conveniently imposes specialist–generalist tradeoffs, in which an increase in performance at one temperature reduces performance at another temperature. As in their original model (Lynch and Gabriel 1987), values of z1 and z2 of an individual were determined by genotypic and environmental factors: zi = gi + ei ,

No plasticity Plasticity in z1 or z2

(a) 0.5

141

0.3 0.2 0.1 0.0 –3

–2

–1

0

1

2

3

4

5

Temperature (°C) Figure 5.8 Developmental acclimation of the thermal optimum (a) or performance breadth (b) enhances the performance of an organism in a heterogeneous environment. Genotypic values of the thermal optimum and performance breadth (g1 and g2 , respectively) interact with the environmental temperature to produce phenotypic values (z1 and z2 , respectively). Environmental temperatures are scaled relative to the mean (i.e., mean = 0). In these examples, the developmental temperature exceeds the mean temperature by 2◦ C. The arrow in each plot shows the increase in survivorship derived from acclimation to the developmental temperature. Adapted from Gabriel (1999) in The Ecology and Evolution of Inducible Defenses (Harvell, C. D. and R. Tollrian, eds.) by permission from Princeton University Press. © 1999 Princeton University Press.

ﬁtness of an organism by reducing the need for a wide performance breadth. The optimal degree of acclimation depends on the spatial and temporal variance of environmental temperature. Selection favors a greater capacity for acclimation as the spatial variation within generations or the temporal variation among generations increases. In contrast, increasing the temporal variation within generations lessens the selective advantage of acclimation.

142

T H E R M A L A D A P TAT I O N

ri

Boundary z*i – gi

In a sense, the temporal variation experienced by an individual deﬁnes the stochasticity of the environment. A more stochastic (or less predictable) environment favors a relatively wide performance breadth instead of a developmental shift in the thermal optimum. Therefore, the greatest capacity for developmental acclimation should evolve in environments in which thermal heterogeneity among generations greatly exceeds that within generations. Acclimation of the performance breadth yields little selective advantage over acclimation of the thermal optimum alone (Fig. 5.10). Given the potential for the thermal optimum to acclimate, the optimal performance breadth depends strictly on the variance of temperature within generations. A relatively constant (or predictable) environment favors the development of a specialist for the mean temperature. Even in the most heterogeneous environments, acclimation of the thermal optimum confers a greater selective advantage than acclimation of the performance breadth. This result follows from the assumption that the maximal performance of a specialist exceeds that of a generalist. If an organism can match its thermal optimum to its current environment, why accept the cost of being a generalist? Through developmental acclimation, an organism can perform over a wide range of temperatures without sacriﬁcing performance at the optimal temperature. Unless some other constraint exists, the

Maximal selective advantage

Figure 5.9 In the model by Gabriel and Lynch, a nonlinear reaction norm described a bounded relationship between the optimal shift in the phenotype (zi∗ − gi ) and the acclimation response (ri ). Adapted from Gabriel and Lynch (1992) with permission from Blackwell Publishing.

3.0 w(r1)/w0

2.5 2.0 1.5

w(r2)/w0

1.0 0.5

0

5 (Vft + Vftb)/Vftw

10

Figure 5.10 The selective advantage of developmental acclimation of the thermal optimum (z1 ) or performance breadth (z2 ) depends on several sources of thermal heterogeneity: spatial variation (Vφt ), temporal variation within generations (Vφtw ), and temporal variation among generations (Vφtb ). Solid lines depict relative ﬁtnesses of a plastic genotype to a ﬁxed genotype for zi : w (ri )/w0 . The dotted line represents the selective advantage of a change in the thermal optimum without a correlated change in the performance breadth. The selective advantage increases with an increase in the ratio of the variation experienced among individuals to the variation experienced within individuals, (Vφt + Vφtb )/Vφtw . These selective advantages are for unbounded reaction norms for z1 and z2 . Adapted from Gabriel and Lynch (1992) with permission from Blackwell Publishing.

thermal optimum of performance should acclimate more readily than the performance breadth. Gabriel and Lynch formulated their model under the assumption that acclimation does not cost energy

T H E R M A L A C C L I M AT I O N

or reduce survivorship. Therefore, let us brieﬂy consider how such costs would alter the predictions of their model. Because developmental acclimation enables organisms to specialize for local temperatures, the ﬁtness of this strategy must outweigh that of a generalist. Assuming generalization imposes a reduction in maximal performance (see Chapter 3), a cost of acclimation exists at which both acclimation and generalization yield equal ﬁtness (DeWitt and Langerhans 2004). Below this cost, developmental acclimation would be favored over generalization. In this range, the qualitative results of Gabriel and Lynch’s analysis hold unless the cost of acclimation depends on the thermal environment or the acclimation response. Given that generalization imposes a severe cost (Lynch and Gabriel 1987), we should expect useful insights to emerge from this model despite its assumption of no energetic cost. Gabriel and Lynch also assumed that organisms receive accurate cues from their environment. But what if the temperature during development poorly predicts the temperature later in life? Would such unpredictability affect the evolution of developmental acclimation? Moran (1992) addressed this very question by modeling ﬁtness landscapes in a simple, coarse-grained environment. She assumed an environment that comprised two kinds of patches, which we can think of as hot and cold. An individual disperses randomly after birth and experiences a single patch during its lifetime. A cold-adapted phenotype achieves greater ﬁtness in a cold patch than does a heat-adapted phenotype; the opposite occurs when the two phenotypes encounter hot conditions. A plastic genotype could adopt either phenotype but would suffer a cost if the wrong phenotype were expressed. The optimal strategy depends strongly on the accuracy of environmental cues and the frequency of the two environments. If the accuracy of the cue exceeds the frequency of the more common environment, a genotype capable of developmental acclimation would outperform one that specializes on a single environment. If the cue is less accurate, selection favors a specialist for the more frequent of the two environments. In other words, plastic genotypes must express the correct phenotype more often than nonplastic genotypes would do so by chance alone. If one environment occurs much more frequently than the other, the evolution of

143

developmental acclimation requires (1) extremely accurate cues, (2) a low cost of expressing the wrong phenotype in the more common environment, or (3) a high cost of expressing the wrong phenotype in the less common environment. Thus, developmental acclimation would likely evolve when the frequencies of different thermal patches equal one another, or when the ﬁtness of one phenotype greatly exceeds that of the other phenotype in the rare patch.

5.4.2 Optimal reversible acclimation Compared to developmental acclimation, reversible acclimation offers even greater potential to match physiology to the current environment. To explore this potential, Gabriel (1999) modeled the optimal strategy of reversible acclimation in a simple environment that switched between two states. Although Gabriel deﬁned the environment in general terms, we can think of these states as operative environmental temperatures (T1 and T2 ); they could represent either diel or seasonal extremes depending on the temporal scale we choose. As in previous analyses, the thermal sensitivity of survivorship was modeled as a Gaussian function, deﬁned by z1 and z2 (see Section 5.4.1). The environment shifted abruptly between T1 and T2 , each state lasting a certain period (ts ). Following the shift from T1 to T2 , an organism adapted to T1 would lose some ability to perform. Should the organism alter its performance curve in response to the change in temperature? The beneﬁt of doing so seems obvious, but what about the costs? Gabriel focused on the cost imposed by time lags. As discussed in Section 5.3.2, if a response to thermal change requires time, the temperature might change again before the response can occur. With this in mind, we must compare the performance of two genotypes: one whose performance curve acclimates reversibly (+) and one whose performance curve remains ﬁxed (−). During a switch in temperature, the performance of both the acclimating and nonacclimating genotypes would decrease. After a time lag, however, the performance of the acclimating genotype would rebound. Once the environment switches back to its original temperature, the acclimating genotype suffers another temporary

T H E R M A L A D A P TAT I O N

loss of performance while the nonacclimating genotype regains its original level of performance. The relative ﬁtness of these genotypes (w+ /w− ) depends on the time required for an acclimation response (tr ) and the frequency and magnitude of environmental change:

2

(T −T ) 0.5(ts −tr ) 2 z 1 w+ 2 =e w−

.

(5.4)

Decreasing either the frequency of environmental change (1/ts ) or the lag during acclimation (tr ) enhances the relative ﬁtness of reversible acclimation. Similarly, increasing the magnitude of environmental change (T2 – T1 ) also favors acclimation. Speciﬁcally, the optimal shift in the thermal optimum (z1∗ ) during acclimation would be tr . (5.5) z1∗ = (T2 − T1 ) 1 − 2ts Obviously, acclimation never beneﬁts an organism when the time needed to acclimate (tr ) exceeds the time between thermal switches (ts ). Hence, reversible acclimation should beneﬁt long-lived organisms, which experience seasonal changes in temperature during their life. Furthermore, reversible acclimation would also beneﬁt organisms during diel variation in temperature if acclimation and deacclimation occur rapidly. In contrast to developmental acclimation, reversible acclimation of the performance breadth can provide a signiﬁcant beneﬁt. Two conditions favor reversible acclimation of the performance breadth: (1) when acclimation occurs slowly but the environment changes frequently or (2) when the temperature shifts greatly between periods (Fig. 5.11). The beneﬁts of a greater performance breadth under such conditions might seem obscure, but the key lies in thinking about performance during deacclimation. Acclimation of the thermal optimum causes maladaptation to T1 in exchange for adaptation to T2 ; however, acclimation of the performance breadth confers some performance at T2 while reducing maladaptation to T1 during deacclimation. This aspect of the model has direct implications for adaptation to diel and seasonal cycles (see Fig. 5.11). Diel ﬂuctuations, which occur rapidly relative to physiological responses, should favor acclimation of the performance breadth. Seasonal ﬂuctuations,

1.5 Relative time lag (tr/ts)

144

Diel variation (tropical)

Diel variation (temperate)

1.0

0.5

0.0

Seasonal variation (tropical) 0

Seasonal variation (temperate)

2 4 6 8 10 Magnitude of thermal change (T2 – T1)

12

Figure 5.11 The selective advantage of reversible acclimation depends on the magnitude of thermal change (T2 − T1 ) and the relative time lag for acclimation (tr /ts ). The lines depict isoclines of equal ﬁtness resulting from acclimation of the thermal optimum (z1 ) and acclimation of the performance breadth (z2 ). Above the line, acclimation of the performance breadth yields greater ﬁtness than acclimation of the thermal optimum; below the line, the reverse condition holds. From left to right, the three isoclines represent increasing performance breadths prior to acclimation (g2 equals 1, 4, and 9, respectively). The parameter values at each of the four corners of the plot correspond to certain temporal patterns of variation in either temperate or tropical environments (see Chapter 2). Adapted from Gabriel (1999) in The Ecology and Evolution of Inducible Defenses (Harvell, C. D. and R. Tollrian, eds.) by permission from Princeton University Press. © 1999 Princeton University Press.

which are more prolonged than diel ﬂuctuations, should favor acclimation of the thermal optimum. This prediction should hold regardless of the environment in which the organism has evolved (see Fig. 5.11). As in his previous model (Section 5.4.1), Gabriel assumed reversible acclimation imposed no cost in terms of energy or survivorship. We can easily infer how such a cost would affect the predictions of his model. Figure 5.12 depicts the ﬁtnesses of various strategies of reversible acclimation. Because acclimation imposes no cost in this analysis, the ﬁtness of an acclimating genotype always exceeds that of a nonacclimating genotype. A direct cost of acclimation would shift the ﬁtness curves for acclimating genotypes downward, without affecting the curve for the nonacclimating genotype. As a consequence, selection would favor a genotype that did not acclimate in relatively stable environments. All other things being equal, selection for

T H E R M A L A C C L I M AT I O N 3

0.5 Acclimation of z1 Acclimation of z2 No acclimation

z1 and z2 (complete) Relative fitness

Fitness

0.4 0.3 0.2 0.1 0

145

0

2 4 6 8 Magnitude of thermal change (T2 – T1)

10

Figure 5.12 The selective advantage of reversible acclimation depends on the magnitude of thermal change (T2 – T1 ). Each line shows this dependence for a genotype that can change either its thermal optimum (z1 ) or its performance breadth (z2 ), or neither parameter. Other parameters were set as follows: ts = 0.4, tr = 0.1, and g2 = 1. Adapted from Gabriel (1999) in The Ecology and Evolution of Inducible Defenses (Harvell, C. D. and R. Tollrian, eds.) by permission from Princeton University Press. © 1999 Princeton University Press.

costly reversible acclimation would require a greater magnitude of thermal change or a shorter time lag (Padilla and Adolph 1996). The predictability of temperature also affects the optimal strategy of reversible acclimation. Gabriel (2005) extended his model of reversible acclimation to consider the effect of unpredictable magnitudes of thermal change. To introduce unpredictability, he assumed that T2 was drawn from a distribution of temperatures; thus, the organism knows the environment has changed but does not know by how much. If the organism can detect a change in temperature but cannot predict the magnitude of this change, the optimal shift in the thermal optimum corresponds to the long-term mean of thermal change (i.e., mean T2 − mean T1 ). Acclimation of the performance breadth in addition to acclimation of the thermal optimum helps to compensate for the stochasticity of thermal change (Fig. 5.13). The optimal performance breadth after acclimation scales proportional to the variance of T2 . Even with the potential for acclimation, stochasticity reduces the ﬁtness of a genotype in a heterogeneous environment.

2.5

2

1.5

z1 and z2 (incomplete) z1 (complete)

1 2 3 4 Coefficient of variation for T2

Figure 5.13 In a stochastic environment, the selective advantage of reversible acclimation of the thermal optimum (z1 ) or the performance breadth (z2 ) depends on whether the organism can determine the magnitude of thermal change (complete information) or only the occurrence of thermal change (incomplete information). Fitnesses are scaled relative to the ﬁtness of a genotype whose thermal optimum acclimates with incomplete information. The three lines represent the following genotypes: (1) acclimation of the thermal optimum with complete information, (2) acclimation of the thermal optimum and the performance breadth with incomplete information, and (3) acclimation of the thermal optimum and the performance breadth with complete information. Adapted from Gabriel (2005) with permission from Blackwell Publishing.

5.4.3 Relaxing assumptions about ﬁtness Although the models presented in this chapter describe ﬁtness as the product of performance over time, we could easily envision a scenario in which ﬁtness depends on the sum of performance. In Chapter 3, we saw how the relationship between performance and ﬁtness affects the evolution of thermal sensitivities. Recall that most patterns of thermal heterogeneity favor the evolution of thermal specialists when performance contributes additively to ﬁtness; generalists are favored only when the variation among generations exceeds the variation within generations (Gilchrist 1995). No formal model of optimal acclimation has addressed this scenario, but we can intuit the behavior of such a model. Acclimation of the thermal optimum relaxes the selective pressure imposed by thermal heterogeneity among generations. Therefore, acclimation would always favor specialists when performance contributes additively to ﬁtness. Consequently, the selective advantage of developmental acclimation would not depend on the thermal heterogeneity

146

T H E R M A L A D A P TAT I O N

within generations. Similarly, reversible acclimation of the performance breadth would not beneﬁt an organism. Of course, both of these conclusions hinge on the existence of specialist–generalist tradeoffs.

5.5 Evidence of optimal acclimation The models described in the preceding section provide testable predictions about the acclimation of performance curves. First, thermal optima should acclimate more often than performance breadths, particularly during prolonged changes in temperature (i.e., seasonal or intergenerational changes). Conversely, performance breadth should acclimate primarily in response to rapid changes in temperature, such as diel cycles. Second, organisms in variable environments should acclimate more readily than organisms in stable environments. In this section, I review the experimental evidence for each of these hypotheses. Keep in mind that most of the experiments were not designed explicitly to test the theory of optimal acclimation.

5.5.1 Does the thermal optimum acclimate more than the performance breadth? Although the theory suggests the thermal optimum of performance should acclimate readily in variable environments, studies of plants and animals indicate that thermal optima rarely respond to changes in environmental temperature (see Tables 5.1 and 5.2). As discussed in Section 5.2, plants generally exhibit an optimal temperature for development, rather than plasticity of their thermal optima (but see Mooney 1980; Slatyer 1977; Smith and Hadley 1974). Some species of animals undergo acclimation of locomotor performance consistent with the theory. For example, a few species of ﬁsh (Fry and Hart 1948; Grifﬁths and Alderdice 1972; Johnson and Bennett 1995; Larimore and Duever 1968; Seebacher et al. 2005) and one crocodilian (Glanville and Seebacher 2006) adjusted their thermal optima for swimming following abrupt changes in temperature. Similarly, the thermal optimum of ﬁltration by oysters scaled proportional to acclimation temperature (Newell et al. 1977). These examples merely

prove exceptions to the rule. The majority of studies revealed no evidence for the optimal acclimation of thermal optima (see Table 5.2).

5.5.2 Do organisms from variable environments acclimate more than organisms from stable environments? In principle, we could answer this question through an experimental or a comparative approach. A powerful experimental approach involves natural selection in the laboratory (see Section 1.5.2). One could establish replicated populations and force them to evolve in constant and ﬂuctuating environments. After some period, one would compare the acclimation capacities of individuals from these populations. Unfortunately, no researcher has conducted a selection experiment designed to address the evolution of acclimation. For this reason, we must rely solely on comparisons of acclimation capacities among natural populations. An ideal comparative approach involves a quantitative analysis of acclimation capacity in a phylogenetic context (see Section 1.5.3). The populations or species in this analysis would span a range of thermal environments, perhaps distributed over latitudinal or altitudinal gradients. Unfortunately, comparative studies of acclimation generally lack the replication required for quantitative phylogenetic analysis. Therefore, we must rely on qualitative comparisons of acclimation capacity between genotypes from different thermal environments. Comparative studies of developmental acclimation provide virtually no support for the notion that developmental acclimation evolves according to thermal heterogeneity. Most of the relevant data come from studies of photosynthetic rates (Table 5.3). In an early review of this subject, Berry and Björkman (1980) concluded the acclimation capacity of a plant species corresponds to its selective environment. Among six species they considered, three patterns were observed: (1) two species from cold environments performed poorly after developing at high temperatures, (2) one species from a hot environment performed poorly after developing at low temperatures, and (3) three species from variable environments acclimated well to high and low temperatures (see Fig 5.2). More

T H E R M A L A C C L I M AT I O N

147

Table 5.3 Comparative studies of the thermal acclimation of photosynthetic rate (adjusted for the area or mass of leaves). For each study, I note patterns that are relevant to current optimality models. I also note whether the experiment revealed evidence of an optimal developmental temperature (ODT).

Species

Cline

Acclimation temperatures

Test temperatures

Acer saccharum

Latitude

27/18, 31/18

11, 16, 21, 26, 31, 36

Atriplex lentiformis

Desert vs. coastal

23/18, 43/30

10–46◦ C in 2◦ C increments

Australian trees

Latitude

22◦ C:14◦ C and a stochastic cycle with the same mean temperature

12, 17, 22, 27, 32

Australian trees

Latitude

14/6, 19/11, 22/14, 25/17, 30/22

10, 14, 18, 20, 22, 24, 26, 30

Eucalyptus pauciﬂorac

Altitude

Heliotropium curassavicum

Desert vs. coastal

8/4, 15/10, 21/16, 27/22, 33/28 25/15, 40/26

10, 15, 20, 25, 30, 35, 40 20, 25, 30, 35, 40, 45

Ledum groenlandicum

Latitude and longitude Altitude

15/10, 30/25

15, 20, 25, 30, 35

12/4, 21/10, 32/21

10, 21, 32, 38, 43

Oxyria digyna

Pattern

ODT

Source

Plants performed better when they had developed at higher temperatures. Topt was related to acclimation temperature in desert plants, but development at low temperature resulted in maximal performance of coastal plants. Temperate species underwent acclimation of performance breadth, whereas tropical species did not; however, the thermal breadth of photosynthesis was greater in the less variable and more predictable treatment. Plasticity of the thermal optimum was greater in tropical species than in temperature species; performance breadth did not acclimate. Topt was related to acclimation temperature in all populations. Topt was related to acclimation temperature in both desert and coastal plants. Topt was related to acclimation temperature in all four populations. Topt was related to acclimation temperature in alpine populations, but Arctic populations performed best when they had developed at low temperatures.

Yes

Gunderson et al. (2000) Pearcy (1977)

Yes/no

No

Cunningham and Read (2003b)

Yes

Cunningham and Read (2002)

No

Slatyer (1977)

No

Mooney (1980)

No

Smith and Hadley (1974) Billings et al. (1971)

Yes/no

Continued

148

T H E R M A L A D A P TAT I O N

Table 5.3 Continued

Species

Cline

Acclimation temperatures

Test temperatures

Plantago major, P. euryphylla, and P. lanceolata

Altitude

13, 20, 27

6, 13, 20, 27, 34

Valonia utricularis

Latitude

15, 18, 20, 25

5, 10, 15, 20, 25, 30, 35

Valonia utricularis

Latitude

≤ 20, 25, 30

≤ 20, 25, 30

Pattern

ODT

Source

Alpine plants performed better at low temperatures (13–20◦ C), regardless of their developmental temperature. Lowland plants that developed at 13◦ C performed better at low temperatures (6–13◦ C) than plants that developed at 20◦ C or 27◦ C. Topt was related to acclimation temperature in algae from the Indian Ocean, but Topt did not acclimate in Mediterranean algae. Under chronic exposure to low temperatures, temperate algae maintained high rates of photosynthesis but tropical algae did not. The opposite pattern was observed when algae were exposed to high temperatures

Yes

Atkin et al. (2006b)

Yes

Eggert et al. (2003b)

Yes

Eggert et al. (2006)

Notes: Topt , thermal optimum for performance.

rigorous comparisons suggest a very different view: little correspondence exists between thermal heterogeneity and acclimation capacity. For several species, populations exhibited strong acclimation of thermal optima regardless of the thermal heterogeneity of their native environment (Mooney 1980; Slatyer 1977; Smith and Hadley 1974). Moreover, genotypes from variable environments sometimes exhibited less capacity for acclimation than genotypes from stable environments. In a comprehensive case study, Cunningham and Read (2002) compared the acclimation of

photosynthetic rate in tropical and temperate trees of Australia. These investigators raised eight species of trees in ﬁve ﬂuctuating environments and compared their photosynthetic performance. Contrary to the theory, the thermal optimum acclimated more in tropical species than in temperature species. The performance breadth did not acclimate in any species. These investigators also grew the same eight species under predictable and unpredictable thermal cycles; both treatments had a mean diel cycle of 22◦ C : 14◦ C, but one treatment varied randomly between 17◦ C : 9◦ C and 27◦ C : 19◦ C on a daily basis.

T H E R M A L A C C L I M AT I O N

Consistent with the theory, temperate species had wider performance breadths and underwent greater acclimation of performance breadths than did tropical species. But narrower performance breadths were expressed in the more variable environment! Thermal optima did not acclimate but were 2–3◦ C higher in tropical species than in temperate species. Despite the weak support for current models, at least one key assumption appears valid for these trees: a wider performance breadth resulted in a lower maximal performance. Although rare, comparative studies of reversible acclimation have also produced patterns that contradict the theory. Two examples are particularly relevant. In the ﬁrst example, Cunningham and Read (2003b) followed up on their study of developmental acclimation (see above) by exposing temperate and tropical seedlings to progressively lower temperatures; they started at 30◦ C and ended at 14◦ C, giving seedlings 14 days to acclimate before each decrease in temperature. In most species, the thermal optimum for photosynthesis decreased in response to the decrease in temperature. However, both tropical and temperate species acclimated similarly. In the second example, Deere and Chown (2006) exposed ﬁve species of mites—three marine and two terrestrial—to a range of constant temperatures for 7 days. After this period, they compared the thermal sensitivities of walking speed among mites from different thermal treatments. For four of the ﬁve species, thermal optima and performance breadths shifted little during acclimation. Instead, thermal treatments tended to enhance or impair performance across a wide range of temperatures, leading to variation in mean performance among groups. Clearly, both examples contradict the prediction that reversible acclimation will correspond to the expected change in environmental temperature. Finally, most comparative studies of thermal hardening fail to support the current theory (Table 5.4). Although heat and cold tolerances acclimate readily in many species, the thermal heterogeneity of the selective environment seems to bear little on the capacity for acclimation. A striking exception was documented for D. buzzatii along an altitudinal cline (Sorenson and Loeschcke 2002). Flies from high and low altitudes exhibited

149

distinct diel cycles of heat tolerance. A subsequent experiment conﬁrmed that photoperiod served as a cue for the timing of acclimation; by artiﬁcially coupling photoperiod and temperature, the researchers reversed the diel cycle of heat tolerance in ﬂies from a low altitude, creating ﬂies that tolerated heat better during a cool photophase instead of a warm scotophase.

5.6 Constraints on the evolution of acclimation In general, current models of optimal acclimation predict observed patterns poorly. First, the thermal optimumforperformancerarelyacclimates, whereas thermal tolerances often acclimate during exposure to extreme temperatures. Second, the capacity for acclimation rarely corresponds to the thermal heterogeneity of the selective environment. Why do current optimality models yield such poor predictions about thermal physiology? As with any optimality model, the models described in Section 5.4 focus on functional constraints and ignore genetic constraints, such as additive genetic variance, pleiotropy between traits, and gene ﬂow among populations. In Chapter 3, we saw how genetic variances and gene ﬂow inﬂuence the evolution of performance curves. Now, we shall draw on quantitative genetic models to see how these factors might inﬂuence the evolution of thermal acclimation.

5.6.1 Genetic variance and covariance To explore the potential impacts of additive genetic variance and covariance on the evolution of thermal acclimation, we can use simple quantitative genetic models of phenotypic plasticity. In Chapter 1, we discussed how quantitative genetics models help us to understand constraints on the evolution of reaction norms. If we treat thermal acclimation as a form of phenotypic plasticity, these models can deﬁne constraints on the evolution of acclimation responses. For example, imagine a reaction norm describing how the mean environmental temperature relates to the thermal optimum of performance. The expression of the thermal optimum in multiple environments could be modeled as if it were multiple traits evolving in a single environment

150

T H E R M A L A D A P TAT I O N

Table 5.4 Comparative studies of the acclimation of thermal tolerance in ectotherms

Species

Thermal cline

Acclimation temperatures

Acclimation period

Measure of tolerance

Effect of treatment

Crustacea

Tidal depth

18–32

30–60 min

CTmax

Drosophila buzzatii Drosophila buzzatii

Altitude

39

1 hour

Latitude/other

38

60 min

Drosophila melanogaster Drosophila melanogaster

Latitude

11

3–4 weeks

Knockdown time Survival at 42◦ C for 100 min LT50 at 0◦ C

CTmax of all species increased in winter; CTmax of most species did not acclimate in summer. No acclimation

Latitude

12, 21, 31

Larval period

Recovery from 16 hours at 0◦ C

Drosophila melanogaster

Latitude

4

2 hours

Drosophila melanogaster

Latitude

18, 25

Larval period

Drosophila melanogaster

Latitude

13, 25, 29

4 days

Drosophila similans

Latitude

13, 25, 29

4 days

Drosophila spp.

Latitude

35–38

30 min

Survival at –2◦ C for 23 hours and –5◦ C for 50 min Survival at –2◦ C for 135 min and –5◦ C for 45 min Survival at 39◦ C for 28 min Survival at 39◦ C for 14 min Heat-shock recovery

Fundulus heterclitus Halozetes spp. and Podacarus auberti

Latitude

2, 7, 13, 22, 27, 32, 34 0, 5, 10, 15

≥ 21 days

CTmin / CTmax

7 days

CTmin / CTmax

Orchestia gammarellus

Latitude

5, 10, 15, 20

10 days

CTmax

Petrolisthes spp.

Latitude

sea temperature vs. 10◦ C above sea temperature

Marine vs. terrestrial

CTmax (heart failure)

All populations exhibited similar heat hardening. LT50 of all populations increased sevenfold. Both temperate and tropical populations exhibited strong acclimation, but plasticity was slightly greater for temperature populations. Temperate and tropical populations acclimated similarly.

Source Hopkin et al. (2006) Sorensen et al. (2001) Krebs and Loeschcke (1995) Bubliy et al. (2002) Ayrinhac et al. (2004)

Hoffmann and Watson (1993)

Temperate and tropical populations acclimated similarly.

Hoffmann and Watson (1993)

Temperate and tropical populations acclimated similarly. Temperate and tropical populations acclimated similarly. Flies from low latitude underwent heat hardening, but ﬂies from high latitude were detrimentally affected. Both populations acclimated equally. Both CTmin and CTmax acclimated more in marine species, but only changes in CTmax were in the predicted direction. CTmax of both populations increased with increasing acclimation temperature. CTmin acclimated more and CTmax acclimated less in crabs from lower latitudes.

Hoffmann and Watson (1993) Hoffmann and Watson (1993) Garbuz et al. (2003)

Fangue et al. (2006) Deere et al. (2006)

Gaston and Spicer (1998) Stillman (2003) Continued

T H E R M A L A C C L I M AT I O N

151

Table 5.4 Continued

Species

Thermal cline

Acclimation temperatures

Acclimation period

Measure of tolerance

Effect of treatment

Sarcophagidae

Latitude

0, 25, 40

2 hours

Acclimation of heat and cold tolerances was greater for ﬂies from higher latitudes.

Chen et al. (1990)

Sceloporus spp.

Latitude

1 month

CTmin of all populations decreased in both treatments.

Tsuji (1988)

Tegula spp.

Tidal depth

35/16 at 12L:12D, 35/10 at 8L:16D 14, 22

Survival at –10◦ C or 45◦ C until adulthood CTmin

CTmin / CTmax (heart failure)

Acclimation was greater for subtidal species than for intertidal species.

Stenseng et al. (2005)

15–19 days

Source

Notes:Topt , thermal optimum for performance; CTmin , critical thermal minimum; CTmax , critical thermal maximum; LT50 , the time required for 50% of the individuals in a sample to die from exposure to a temperature. Line 1 females Line 1 males Line 2 females Line 2 males

1.0 Difference in survival between selection and control lines

(see Section 1.4.2). A lack of genetic variance and the presence of genetic covariances will slow adaptation of the reaction norm. Nevertheless, the optimal strategy of acclimation should evolve eventually as long as the population meets two conditions (van Tienderen 1991; Via 1987, 1994). First, some additive genetic variance for acclimation must exist. Second, the acclimation response to one temperature cannot be perfectly correlated (genetically) with the acclimation response to another temperature. If responses at different temperatures were perfectly correlated, the shape of the reaction norm would not be free to evolve toward the optimum. Very little information exists about the genetic variances of acclimation responses (reviewed by Kingsolver and Huey 1998). What we do know pertains mainly to heat hardening by drosophilids. The capacity for heat hardening varies among isofemale lines in some populations (Krebs and Feder 1997) but not in others (Krebs and Loeschcke 1994b; Loeschcke et al. 1994). In Drosophila buzzatii, artiﬁcial selection led to an increase in heat hardening over 10 generations (Fig. 5.14). From Chapter 1, recall that the magnitude of this response to selection reﬂects the heritability of heat hardening in this population. With respect to performance curves, we know virtually nothing about the genetic variance of thermal acclimation. Genetic correlations resulting from pleiotropy would limit thermal adaptation in heterogeneous

0.8 0.6 0.4 0.2 0

–0.2 0

1

2

3 4 5 6 7 8 9 Generations of selection

10 11

Figure 5.14 Artiﬁcial selection led to an increase in heat hardening over 10 generations. In each generation, ﬂies (Drosophila buzzatii) were exposed to 39◦ C for 90 min after a previous exposure to heat. Individuals of selection lines were mated based on their ability to resist heat after acclimation, whereas those from the control lines were mated randomly with respect to this trait. Reproduced from Krebs and Loeschcke (1996) with permission from the Genetics Society of America.

environments. Consider this problem in the context of Gabriel and Lynch’s model of developmental acclimation (Section 5.4.1). The thermal optimum for performance depended on a genetic value (g1 ), an environmental value (e1 ), and an acclimation response (r1 ). The acclimation response (r1 ) was limited by a boundary (b). Now imagine the boundary for acclimation depends on the genetic value for the

152

T H E R M A L A D A P TAT I O N

thermal optimum. Such a relationship constitutes a genetic correlation that constrains the evolution of acclimation. Although genetic correlations of this nature have not been documented, Stillman (2003) discovered a phenotypic correlation consistent with this idea. Among species of porcelain crabs, species that possessed higher critical thermal minima and lower critical thermal maxima exhibited greater capacities for acclimation (Fig. 5.15). This pattern might reﬂect an underlying genetic correlation between thermal sensitivity and thermal acclimation. Alternatively, the phenomenon could

Critical thermal maximum (°C)

(a) 42 40 38 36 34

*

32 *

30 28

Critical thermal minimum (°C)

(b) 12 * 10 8 *

6 4

* 2

*

0 –2 T°C

Acclimation temperature

T + 10°C

Figure 5.15 The acclimation of critical thermal limits for cardiac function in porcelain crabs (Petrolisthes spp.) depended on the value of the thermal limit prior to acclimation. Lines connect critical thermal maxima (a) or critical thermal minima (b) expressed in thermal treatments separated by 10◦ C (T ◦ C and T +10◦ C). Data are for P. eriomerus (ﬁlled circles), P. cinctipes (open circles), P. hirtipes (ﬁlled squares), and P. gracilis (open squares). From Stillman (2003). Reprinted with permission from AAAS.

reﬂect adaptive divergence between crabs in warm, stable environments of the tropics and those in cool, variable environments of the temperate zone.

5.6.2 Gene ﬂow When temperature varies spatially, gene ﬂow among populations can constrain the evolution of thermal acclimation. At ﬁrst, this statement might seem counterintuitive. After all, acclimation enables an organism to adjust its thermal physiology according to its current environment; therefore, migration within a spatially heterogeneous environment should facilitate the evolution of acclimation rather than constrain it (see Section 5.4.1). But when dispersal occurs in the middle of ontogeny, the developmental environment no longer reﬂects the selective environment, increasing the unpredictability of future environmental conditions. To demonstrate this effect, de Jong (1999) modeled the evolution of plasticity for several types of dispersal. She envisioned a plastic trait under stabilizing selection, whose reaction norm could take on any shape deﬁned by a linear or quadratic function. This phenotype equates to a reaction norm for the thermal optimum of performance (e.g., Topt = aT + b). Irreversible acclimation of the phenotype occurred during the zygotic stage but affected the performance of the organism during the adult stage. Three cases of dispersal were considered (Fig. 5.16). In the ﬁrst case, zygotes dispersed randomly and remained in their environment until reproduction (as in Gavrilets and Scheiner 1993; Via and Lande 1985). Because individuals could predict their selective environment under these conditions, the reaction norm in each population converged on the global optimum (i.e., the thermal optimum expressed by a genotype would equal the temperature of its environment). In the second case, zygotes dispersed prior to acclimation and then dispersed again as adults; the dispersal of zygotes determined the developmental environment, but the dispersal of adults determined the selective environment (as in Scheiner 1998). Under these conditions, the reaction norm that evolved in each population differed from the global reaction norm (Fig. 5.17). In the third case, zygotes developed in the same environment as their parents and then dispersed as adults. Again, dispersal after

T H E R M A L A C C L I M AT I O N

(a)

(a)

153

(c) Local zygotes

Mating pool

Mating pool

Selection

Selection

Selection

Adult migration

Adult migration

Development

Development

Development

Zygote disperal

Zygote disperal

Zygote pool

Zygote pool

Local mating

Local zygotes Local mating

Thermal optimum

Figure 5.16 Three distinct life cycles used to model selection and migration. (a) Random dispersal of zygotes, development and selection in the same habitat, and random mating among habitats. (b) Random dispersal of zygotes, development to adulthood, migration among habitats, selection in the adult habitat, and random mating among habitats. (c) No dispersal of zygotes, development to adulthood, migration among habitats, selection in the adult habitat, random mating within each habitat. Reproduced from de Jong (1999) with permission from Blackwell Publishing.

Environmental temperature Figure 5.17 When the dispersal of zygotes and the dispersal of adults disconnect the developmental environment from the selective environment (Fig. 5.16b), the evolved reaction norms (solid lines) deviate from the globally optimal reaction norm (dashed line). A circle marks the temperature of the local environment in which each reaction norm evolved. Each reaction norm intersects the globally optimal reaction norm at the temperature of the genotype’s local environment. Adapted from de Jong (1999) with permission from Blackwell Publishing.

development reduced the predictability of the selective environment. And once again, reaction norms deviated from the global optimum. Because these results hold only under weak selection, we must assume the performances breadth exceeds a certain width to apply this model to the evolution of thermal acclimation. Gene ﬂow also affects the evolution of acclimation when temporal heterogeneity varies geographically. Imagine a species consisting of populations distributed along a latitudinal cline. Without gene ﬂow, the optimal strategy of developmental or reversible acclimation depends on the thermal heterogeneity experienced by a population, not that experienced by the species as a whole (see Section 5.4). To apply current optimality models to metapopulations, we must assume individuals in all localities experience the same temporal variation in temperature. Despite this simplifying assumption, seasonality clearly differs radically across the range of a widespread species. Based on intuition, the optimal strategy of

154

T H E R M A L A D A P TAT I O N

acclimation should depend on both geographic variation in seasonality and the pattern of migration among sites. Unfortunately, no one has formally modeled this phenomenon. By generalizing the models discussed in Section 3.9, we could gain valuable insights about the effects of gene ﬂow on the evolution of thermal acclimation.

5.7 Toward ecological relevance In part, our poor understanding of thermal acclimation stems from a common disregard for the complexity of natural environments. Consider Gabriel’s model of optimal acclimation (Section 5.4.2). In this model, the mean environmental temperature switches abruptly between two states. Moreover, consider the typical experimental design for the study of acclimation (see, for example, Fig. 5.1), in which a researcher exposes organisms to an abrupt shift from one constant temperature to another and subsequently records changes in thermal physiology. This standard design ignores natural variation in environmental temperature. In contrast to theoretical and experimental environments, natural environments change both continuously and unpredictably. Continuous variation in temperature ameliorates the loss of performance during time lags. For example, seasonal change occurs gradually and follows a predictable cycle. Although heat waves and cold snaps add a stochastic element to this cycle, variation among seasons generally exceeds variation within seasons. Two conclusions must follow. First, current models overestimate the loss of performance during acclimation. In other words, the gradual change of natural environments affords a gradual response by an organism. Under gradual thermal change, a given time lag would prove less costly than it would under abrupt thermal change. Second, experimenters have generally exposed organisms to unnatural environmental cues for acclimation. When one manipulates temperature while holding other environmental variables constant, one breaks the natural covariation that organisms rely on to trigger beneﬁcial acclimation. But this natural covariation plays an important role during acclimation. As mentioned above, Sorenson and Loeschcke (2002) showed that ﬂies (D. buzzatii) used the diel cycle of

light to coordinate expression of heat-shock proteins with thermal stress. In this species, light truly supersedes temperature as a cue. Similarly, photoperiod serves as a more reliable cue for seasonal changes in temperature, given the stochasticity of environmental temperature within seasons (see Chapter 2). Because patterns of thermal acclimation depend on other environmental variables (e.g., Hernandez et al. 2006; Licht 1968), the separation of thermal and photoperiodic cues during an experiment could introduce artifacts in the data, leading to erroneous conclusions about adaptation. To conduct ecologically relevant studies of thermal acclimation, researchers must mimic the variability and stochasticity of natural environments. Only two experiments have included such treatments. Schaefer and Ryan (2006) raised zebra ﬁsh (Danio rerio) in constant, variable, and stochastic environments and then compared their critical thermal maxima. Mean temperature was either 24◦ C or 28◦ C, with a daily range of 0◦ C, 4◦ C, or 6◦ C. Not surprisingly, ﬁsh from hotter or more variable environments tolerated higher temperatures. However, ﬁsh that experienced stochastic variation tolerated heat as well as those that experienced predictable variation. This result suggests that the absolute range of temperatures during development inﬂuences acclimation more than the predictability of this range. In another experiment, Niehaus and colleagues exposed ﬁeld crickets (Gryllus pennsylvanicus) to predictable and stochastic changes in temperature (A. C. Niehaus, R. S. Wilson, J. J. Storm, and M. J. Angilletta; unpublished observations). Temperatures of the predictable treatments decreased linearly over time, while those of the stochastic treatment decreased chaotically. The stochastic treatment mimicked the seasonal decline in the temperature of the natural environment (Fig. 5.18). Interestingly, thermal sensitivities of feeding and locomotion failed to acclimate during the experiment, regardless of the thermal treatment. Possibly, the gradual change in temperature provided less information about the beneﬁts of acclimation than an abrupt change in temperature would have provided. Unfortunately, the photoperiod remained constant during the course of the experiment, eliminating a more reliable cue for the seasonal decline in temperature. Although simulating stochastic environments

T H E R M A L A C C L I M AT I O N

(a)

155

40

Temperature (˚C)

30 20 10 0 –10 –20 –30 August

September

October

November

December

January

Time of year

(b) 35

Temperature (°C)

30 25 20 15 10 5 0 0

5

10

15

20

25

30

Day of acclimation

requires more effort than simulating constant ones, these kinds of experiments will likely yield new insights and raise important questions.

5.8 Conclusions A coherent theory of thermal acclimation only recently emerged from efforts to understand the evolution of phenotypic plasticity. This theory consists primarily of optimality models of developmental or reversible acclimation in ﬂuctuating environments. These models represent logical extensions of Lynch and Gabriel’s classic model of the evolution of performance curves, discussed in Chapter 3. While Lynch and Gabriel’s model has had some success in

35

Figure 5.18 Natural temperatures experienced by ﬁeld crickets (a) were used to design a stochastic thermal treatment (b) in an acclimation study (A.C. Niehaus, R.S. Wilson, J.J. Storm, and M.J. Angilletta, unpublished observations). In both plots, upper and lower lines represent maximal and minimal temperatures for each day. Data for natural temperatures were sampled from the year 2000 in Terre Haute, IN, USA.

explaining thermal sensitivities, extensions of this model have had very little success in explaining the acclimation of thermal sensitivities. Contrary to the theory, acclimation of the thermal optimum rarely occurs in laboratory experiments. Additionally, a genotype’s capacity for acclimation rarely correlates with the magnitude or predictability of thermal heterogeneity in its natural environment. Rather, acclimation usually produces variation in performance curves within populations that resembles that produced by evolutionary divergence among populations. As with the evolution of thermal sensitivity, thermal acclimation can lead to phenotypes best described as masters of all temperatures. From this observation, we can draw two conclusions: (1)

156

T H E R M A L A D A P TAT I O N

some factor, not considered by the theory, limits the evolution of thermal acclimation and (2) constraints on thermal acclimation have resulted in optimal temperatures for development. The failure of recent optimality models to account for the observed patterns of acclimation likely reﬂects several problems. First, these models hinge on the assumption that specialist–generalist tradeoffs impose a universal constraint on evolution. We have already discussed the danger of this view in Chapter 3. Second, these models assume sufﬁcient genetic variation for adaptation, yet we know virtually nothing about the genetic variances and covariances of acclimation responses. By reﬂecting back on the previous chapters, we should realize that optimality models have helped us to understand thermoregulatory behaviors far more than they have

helped us to understand thermal sensitivities and their acclimation; the differential success of optimality models might stem from the relative strength of genetic constraints on behavior and physiology. Finally, the theory should ultimately be extended to deal with the complexity of natural environments, such as the interaction between gradual seasonal changes in temperature and sudden stochastic changes. Even with these shortcomings, current optimality models provide a major advance over verbal models such as the beneﬁcial acclimation hypothesis. These quantitative models provide a clear statement of assumptions leading to objective and testable predictions about the evolution of acclimation in a wide range of environments. The explicit nature of their assumptions enables one to generalize these models and advance the theory.

CHAPTER 6

Temperature and the Life History

6.1 The link between performance and the life history In Chapter 3, we needed to translate performance into ﬁtness to predict the evolution of thermal sensitivity. In doing so, we assumed a semelparous organism (i.e., reproducing only once before death), in which either survivorship or fecundity scaled proportional to performance. This choice of life history, whether explicit or not, imposes a particular ﬁtness function. For example, Lynch and Gabriel (1987) used the following function to deﬁne ﬁtness (w) when they modeled optimal performance curves: w = l(α)m(α),

(6.1)

where α equals the age at maturity, l(α) equals the probability of surviving to maturation, and m(α) equals the fecundity at maturation. Performance (P) affected ﬁtness by determining the survivorship: l(α) =

α

P(x).

(6.2)

x=0

Alternatively, Lynch and Gabriel might have chosen a ﬁtness function that enables iteroparous reproduction, such as Euler’s equation: ∞ 1=

e−rx l(x)m(x) dx.

(6.3)

x=α

In this case, the Malthusian parameter (r) would have represented the spread of a genotype in the population, which is a reasonable estimate of ﬁtness in a growing population (see Box 3.3). No biological consideration justiﬁed the use of the simpler function. In fact, by choosing eqn 6.1, Lynch and Gabriel constrained the life history considerably. Still, if they had chosen eqn 6.3, they would have needed to specify the age at maturity and the

age-speciﬁc fecundities. In other words, they would still have constrained the life history substantially. Modeling thermal adaptation of behavior or physiology for speciﬁc life histories serves an important purpose: it yields theoretical insights while keeping the mathematics tractable. But we could just as easily wonder how traits such as age at maturity or age-speciﬁc fecundity evolve in response to thermal heterogeneity. In this chapter, we shall ﬂip the coin and ask the following: assuming a speciﬁc performance curve, how would natural selection shape the life history? Thermal heterogeneity favors the evolution of plastic life histories. In Chapter 3, we saw that temperature sets the capacity for physiological performance. Performances such as energy assimilation constrain the evolution of life-history traits, including the age and the size at maturity, the frequency and magnitude of reproduction, and the size and the number of offspring. These life-history traits deﬁne the parameters of the ﬁtness function—l(x) and m(x). Therefore, reaction norms for life-history traits should evolve as a consequence of the thermal sensitivity of performance. We shall focus on two fundamental questions about the evolution of life-histories. First, how do thermal sensitivities of growth and survivorship affect the optimal reaction norms for age and size at maturity? Second, should organisms in cold environments make many small offspring or a few large ones, relative to organisms in warm environments? To answer each of these questions, we can draw on a rich store of models that has accumulated in recent decades. These models will help us to understand the thermal adaptation of life-history traits in ectotherms. To understand thermal adaptation of the lifehistory, we must consider tradeoffs that constrain the allocation of energy. Indeed, such tradeoffs 157

T H E R M A L A D A P TAT I O N

have formed the core of life-history theory since its inception (reviewed by Roff 2002; Stearns 1992). Individuals allocate their energy to the competing demands of maintenance, growth, and reproduction. Maturation generally marks a pronounced shift in the allocation of energy from maintenance or growth to reproduction. Presumably, natural selection shapes the relative investment in each of these functions, and these allocations produce the lifehistory traits that we observe (Dunham et al. 1989). In practice, theorists rarely focus on the entire life history at once because the mathematics would become very complex. Rather, they partition the life history into interesting yet manageable chunks. For instance, if we wanted to model the evolution age and size at maturity, we would likely focus on the tradeoff between growth and reproduction. The allocation of energy to growth would beneﬁt future reproduction at the expense of current reproduction. Alternatively, if we wanted to understand the evolution of reproductive traits, we should focus on the tradeoff between the size and number of offspring. The allocation of energy to each offspring reduces the potential number of offspring. In either case, the optimal phenotype emerges from the tradeoffs that dictate the phenotypic function (see Chapter 1). Temperature modulates the ﬁtness consequences of these tradeoffs through its inﬂuence on performance. This theme permeates the remainder of this chapter.

6.2 General patterns of age and size at maturity 6.2.1 Thermal plasticity of age and size at maturity At least a decade before the proliferation of lifehistory theory, Ray (1960) discovered that many ectotherms respond similarly to variation in their developmental temperature. By raising 13 species of plants and animals at different temperatures, he observed the same pattern in almost every species: individuals at low temperatures grew relatively slowly, but delayed maturation long enough to outgrow individuals at high temperatures. More than 30 years later, Atkinson further generalized Ray’s

80

Number of species

158

60

40

20

0 Negative

No effect

Positive

Thermal effect on body size

Figure 6.1 The majority of ectothermic species studied in the laboratory attained larger sizes when raised at lower temperatures. Data were summarized by Atkinson (1994).

observation. In a review of 109 experiments, Atkinson (1994) concluded Ray’s relationship between temperature and size was strikingly widespread; 80% of species12 —including animals, plants, protists, and one bacterium—reached larger sizes at lower temperatures (Fig. 6.1). Atkinson went as far as to suggest a “near-universal relationship,” which he later called the temperature–size rule (Atkinson 1996a). Subsequent experiments have supported the generality of this relationship (Fischer and Fiedler 2000, 2001a; Møller et al. 1989; Stillwell and Fox 2005), despite the discovery of additional exceptions (e.g., De Block and Stoks 2003). Most recently, Atkinson and colleagues (2003) conﬁrmed the generality of the temperature–size rule among protists. Without a doubt, the temperature–size rule ranks among the most taxonomically widespread “rules” in biology. 12

Although Atkinson reviewed 109 experiments, some of these experiments involved the same species (e.g., he included seven studies of Drosophila melanogaster). Treating these experiments as independent would ignore the evolutionary history shared by populations of a species (see Chapter 1 and Box 6.1). Therefore, I collapsed these results into a single result for each species, as the results of these experiments generally agreed with one another. For two species, a negative effect of temperature was observed in one or more experiments and mixed effects were observed in another experiment. For these species, I assumed some investigators failed to raise organisms over a wide enough range of temperature to detect a maximal body size.

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

Despite the tendency described as the temperature– size rule, reaction norms for stage-speciﬁc size do vary within and among populations. Two degrees of variation call for our attention. First, the strength of the relationship between temperature and size varies among genotypes, even when all genotypes follow the temperature–size rule. This phenomenon has been documented among isogenetic lines of drosophilids (Loeschcke et al. 1999) and among distinct populations of other insects (Stillwell and Fox 2005). Second, the very nature of the temperature– size relationship can differ among genotypes. For example, three populations of snails diverged in their reaction norms for size at maturity; genotypes in one population followed the temperature–size rule while those in the other populations did not respond to developmental temperature (Dybdahl and Kane 2005). A similar divergence occurred between two populations of butterﬂies in less than 150 years (Fig. 6.2).

159

Likely, a nonlinear function better describes the true form of these thermal reaction norms, such that smaller size results from development at extremely high and low temperatures. For example, several species of drosophilids attained their largest size when developing between 15◦ C and 20◦ C (David et al. 2006; Petavy et al. 1997; Yadav and Singh 2005). Similarly, Daphnia pulex attained its largest length at maturity when developing at an intermediate temperature (Dufresne and Hebert 1998). Moreover, the effect of developmental temperature on the life history depends on other environmental conditions, such as the abundance of food or the risk of predation (Gebhardt and Stearns 1993; Newman 1998; Weetman and Atkinson 2002). Despite these exceptions, the temperature–size rule still describes the thermal plasticity of many genotypes over a wide range of temperatures. Thus, even when reaction norms appear nonlinear, we should wonder why organisms that grow very rapidly would cease growing at a relatively small size.

Adult mass (mg)

(a) 105

6.2.2 Thermal clines in age and size at maturity

100 95 90 85 80

Development time (days)

(b)

Cool

Warm

Cool

Warm

40 38 36 34 32 30 28 26

Figure 6.2 Thermal reaction norms for adult mass (a) and development time (b) diverged between populations of butterﬂies (Pieris rapae) from North Carolina (solid line) and Washington (dashed line) Error bars represent standard errors. Adapted from Kingsolver (2007) with permission from Blackwell Publishing.

Species distributed over broad geographic ranges often exhibit thermal clines in body size. Recent phylogenetic comparative analyses indicate the majority of tetrapods reach larger adult sizes in colder environments (Box 6.1), a phenomenon widely referred to as Bergmann’s rule (Blackburn et al. 1999). The patterns described by Bergmann’s rule parallel those described by the temperature–size rule. Although Bergmann originally considered variation in size among endotherms, the size clines that bear his name seem to occur in many ectotherms as well. Among tetrapods, the majority of amphibians and turtles exhibit Bergmann’s clines along latitudinal or altitudinal gradients (Ashton 2002a; Ashton and Feldman 2003; Measey and Van Dongen 2006; Morrison and Hero 2003b), whereas most squamates exhibit opposing clines (i.e., smaller sizes in colder environments). Among ﬁshes, Bergmann’s clines rarely occur in freshwater species, but occur commonly in anadromous and marine species (Belk and Houston 2002; Blanckenhorn et al. 2006; Gilligan 1991).

160

T H E R M A L A D A P TAT I O N

Box 6.1 Phylogenetic comparative analyses of Bergmann’s clines How did researchers assess the generality of Bergmann’s clines? In general, this task requires a meta-analysis, or an analysis of analyses (see Gurevitch and Hedges 1993). Two meta-analytical approaches were taken: vote counting and correlational analysis. For the vote-counting approach, researchers treated a size cline as discrete trait, assigning a “plus” to all species exhibiting a Bergmann’s cline and a “minus” to all other species. Then, they used a statistical test to see whether the proportion of species with a Bergmann’s cline exceeded the expected proportion of 0.5. When vote counting, some investigators counted only statistically signiﬁcant Bergmann’s clines (e.g., Meiri and Dayan 2003) while others counted all trends, regardless of statistical signiﬁcance (e.g., Ashton et al. 2000). For the correlational approach, researchers calculated the coefﬁcient of the correlation between latitude (or temperature) and size for each species. Then, these coefﬁcients were used to estimate a mean latitudinal (or thermal) effect, weighted for the sample size of each species (see Gurevitch and Hedges 1993). With either approach, one should worry about pseudoreplication resulting from the evolutionary relationships among species (see Chapter 1). This problem compromises a statistical inference when the data possess a strong phylogenetic signal. Some novel efforts have been made to detect phylogenetic signal or account for its inﬂuence on Bergmann’s clines. Ashton (2002b) computed independent contrasts of correlation coefﬁcients and their associated variance. Then, he randomized correlation coefﬁcients within the phylogeny and recalculated the variance of the contrasts; repeating this procedure 1000 times resulted in a null distribution for his test statistic. Because the real variances always fell well within the distribution of simulated variances, Ashton ruled out the presence of a strong phylogenetic signal. To control for phylogeny during meta-analysis, Ashton and Feldman (2003) simulated the evolution of correlation coefﬁcients within a phylogeny, assuming equal branch lengths and an evolutionary process equivalent to Brownian motion. For each simulation, they calculated the proportion of extant species that followed Bergmann’s rule (vote-counting approach) and the mean correlation coefﬁcient (correlational approach). After 1000 simulations, they asked

whether the observed latitudinal effect was more extreme than the expected latitudinal effect caused by their simulated evolution. Keep in mind, these analyses focused on the evolutionary relationships among species but ignored the relationships among populations within species. Addressing the latter issue requires a phylogenetic hypothesis for each species (see, for example, Angilletta et al. 2004b). Phylogenetic analyses of Bergmann’s clines appear robust to several potentially confounding variables, such as the mean body sizes, mean latitudes, or latitudinal ranges of the species. To check for systematic bias in the data, Ashton (2004) computed phylogenetic contrasts of the correlation coefﬁcients and regressed these contrasts against potentially confounding variables. The direction and magnitude of a size cline was unrelated to the mean body size, mean latitude, or latitudinal range. However, larger samples yielded stronger clines in body size, suggesting that statistical power limited the resolution of size clines for some species. Overall, these analyses underscore the reality of size clines and justify our search for adaptive explanations. de Quieroz and Ashton (2004) concluded that a Bergmann’s cline was a deeply ancestral characteristic of tetrapods. They came to this conclusion by mapping Bergmann’s clines onto a phylogenetic hypothesis for the Tetrapoda. Although de Quieroz and Ashton’s argument seems interesting, I am skeptical of their interpretation. The problem lies in equating a Bergmann’s cline with a heritable trait. An organism cannot pass on a geographic cline to its offspring; it can only pass on genes that affect a thermal reaction norm. By mapping size clines as traits, de Quieroz and Ashton implicitly assumed each size cline reﬂected a reaction norm, passed from ancestor to descendant. In which case, the size clines must have resulted from thermal plasticity rather than genetic divergence among populations. But if size clines resulted from genetic divergence among populations, this genetic structure would not necessarily pass to descendents during speciation. Because body size adapts readily to local environments (Fig. 6.2; Partridge and Coyne 1997), this logical ﬂaw likely undermines the main conclusion reached by de Quieroz and Ashton.

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

Mean femur length (mm)

In contrast to vertebrates, arthropods seem to violate Bergmann’s rule more than they follow it, and sometimes they exhibit more complex patterns. In a recent survey of latitudinal clines, only nine of 33 species exhibited larger sizes at higher latitudes (Blanckenhorn and Demont 2004). A survey of altitudinal clines in insects generated more equivocal results: eight species followed Bergmann’s rule, 14 species opposed the rule, 11 species showed no pattern, and two species showed a complex pattern (Dillon et al. 2006). Complex relationships between latitude and size often resemble a saw-tooth pattern, such as the one seen in crickets (Fig. 6.3); complex relationships of a slightly different form have arisen in some vertebrates, such as the common frog (Rana temporaria) in Scandinavia (Laugen et al. 2005). Granted, we have no idea what size clines look like for the vast majority of arthropods, but available evidence indicates that geographic clines in this group contrast the thermal plasticity observed in the laboratory (see Section 6.2.1). Obviously, geographic clines stem from factors other than thermal plasticity. In the eastern fence lizard (Sceloporus undulatus), females from cold environments mature at larger sizes than females from warm environments; nevertheless, body size appears more closely related to latitude than to temperature (Angilletta et al. 2004b). As with most phenotypic phenomena, geographic clines likely

6.8 Univoltine 6.4

6.0

5.6

Bivoltine

1000 1400 1800 2200 2600 Mean annual season length (day-degrees > 13°C) Figure 6.3 Mean body sizes of ground crickets increased with increasing season length over portions of the range, but decreased sharply at the thermal boundary between populations of univoltine and bivoltine individuals. Adapted from Mousseau (1997) with permission from Blackwell Publishing.

161

depend on the interplay of genetic and environmental factors. In some species, we know that genetic factors contribute to size clines because variation among populations persisted in common garden experiments (Blanckenhorn and Demont 2004; Partridge and Coyne 1997; Schutze and Clarke 2008). Natural selection seems a likely cause of genetic divergence along thermal clines because the direction of genetic differences often parallels the direction of thermal plasticity (i.e., animals from colder environments attain larger sizes). Environmental factors can also enhance or diminish thermal clines in body size. This point was nicely illustrated by Warren and colleagues (2006), who studied Drosophila in an outdoor mesocosm. Hot regions of the environment caused greater mortality of larvae, which in turn reduced competition among surviving larvae. The direct effect of temperature on development and the indirect effect on competition counteracted one another perfectly; larvae in hot microenvironments grew to the same size as those in cold microenvironments. Given the multitude of factors that vary geographically, it is no wonder ectotherms more commonly follow the temperature–size rule in artiﬁcial environments than they do in natural environments.

6.2.3 Experimental evolution of age and size at maturity If thermal reaction norms for age and size at maturity resulted from selection, similar patterns should evolve in the controlled environments of research laboratories. Two selection experiments have adressed this point. In the ﬁrst experiment, Partridge and colleagues (1994) maintained replicated populations of Drosophila melanogaster at either 25◦ C or 16.5◦ C. After 5 years, populations at 16.5◦ C had evolved a larger mean size at maturity than had populations at 25◦ C. Santos and colleagues (2006) conducted a similar experiment with Drosophila subobscura, but obtained a different result. After 3.5 years, wing size had not diverged among populations maintained at 13◦ C, 18◦ C and 22◦ C; nevertheless, rapid development evolved within the lines kept at 18◦ C and 22◦ C. Thus, only age at maturity evolved in a way that parallels patterns of thermal plasticity. From these experiments, we

162

T H E R M A L A D A P TAT I O N

might conclude that natural selection maintains the temperature–size rule, if we assume D. subobscura lacked sufﬁcient genetic variation to respond to selection. Although the selection experiments by Santos and colleagues failed to trigger the evolution of body size, an “experiment” in nature has done so. In the late 1970s, D. subobscura spread from Europe to North and South America. In Europe, this species tends to have larger wings in colder environments (Misra and Reeve 1964). Since invading the Americas, D. subobscura has evolved parallel clines in the wing sizes of females (Huey et al. 2000), although the cellular basis of clinal variation differs among continents (Gilchrist et al. 2004). Larger wings in cold environments might have evolved to reduce wing loading during ﬂight or as a correlated response to selection for larger bodies; available data support the former hypothesis for European ﬂies and the latter hypothesis for South American ﬂies (Gilchrist and Huey 2004). The speciﬁc beneﬁts of large size in a cold environment remain unclear for D. subobscura, but might be inferred from recent studies of D. melanogaster. Like its congener, D. melanogaster evolved Bergmann’s clines on more than one continent. When raised in a common environment, both South American and Australian ﬂies from high latitudes developed faster and matured at larger sizes than their counterparts from low latitudes (James et al. 1995, 1997). At least for this species, temperature appears to play a role in convergent evolution because selection experiments led to the evolution of similar variation in size (see above). Furthermore, researchers have compared components of ﬁtness between ﬂies adapted to low and high temperatures (Bochdanovits and de Jong 2003; McCabe and Partridge 1997; Reeve et al. 2000). Flies that evolved at low temperature enjoyed greater longevity and fecundity in a cold environment than did ﬂies that evolved at high temperature. Interestingly, coldadapted ﬂies enjoyed no noticeable beneﬁts in a warm environment. The convergent evolution of body size during selection experiments and biological invasions favors an explanation based on natural selection over one based on genetic drift. Combined with the abundant evidence for the temperature–size rule, these size clines have motivated some researchers

to seek simple explanations based on physical constraints (van der Have 2002). Nevertheless, the existence of genetic variation within and among populations demands an adaptive explanation for these patterns. With this in mind, let us now ponder the obvious question: why does natural selection in cold environments often favor delayed maturation at a large size?

6.3 Optimal reaction norms for age and size at maturity Although thermal clines and thermal plasticity generally take on a common form, we can always identify exceptions (see Fig. 6.2). Hence the term “rule” must be taken with a grain of salt. Importantly, references to the temperature–size rule or Bergmann’s rule focus attention on common phenomena, but provide no real explanations for these phenomena. We should not think that the temperature–size rule predicts or explains anything; it merely describes the existence of a pattern. To understand why this pattern exists, we must turn to quantitative models of evolution. Optimality modeling provides a tool for understanding thermal reaction norms for age and size at maturity. Two very different types of models exist. The ﬁrst type constrains the organism’s rate of growth, such that body size follows a particular trajectory. The second type constrains the energy assimilated by an organism, such that body size results from the allocation of energy to growth. Both types of models share several key features. First, the models focus on tradeoffs between current and future fecundity, mediated by strategies of growth. Second, they assume temperature remains constant throughout the life of an organism; this assumption eliminates the need to worry about the asymmetry of performance curves. Finally, these models assume plasticity occurs without cost or constraint (i.e., complete plasticity, sensu Sibly and Atkinson 1994). In this section, we shall consider an example of each type of optimality model and explore the conditions under which it can explain the temperature–size rule. In doing so, we shall see that the optimal life history depends on the thermal sensitivities of juvenile survivorship, maximal size, and population growth.

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

(b)

(c)

Cold

Age

Rate of mortality

Hot Fecundity

Size in length

(a)

163

smin

Size at maturity

a

Age

Figure 6.4 Hypothetical relationships used to model optimal reaction norms for age and size at maturity (based on Berrigan and Koella 1994). (a) Growth follows the Bertalanffy function, such that initial growth rate increases as the parameter k increases. (b) Fecundity increases linearly with increasing size once the organism reaches a minimal size (smin ). (c) The rate of mortality remains constant during the juvenile and adult stages, but can differ between stages; the example shown here depicts a greater rate of mortality for juveniles.

6.3.1 A comparison of two modeling approaches Berrigan and Koella (1994) modeled optimal reaction norms relating growth rate to age and size at maturity. Because growth rate depends on temperature, their model nicely illustrates the selective pressures that thermal heterogeneity can impose on the life history. Berrigan and Koella started by deﬁning the ﬁtness of an organism as the net reproductive rate: R0 =

∞ l(x)m(x) dx.

(6.4)

α

In doing so, they assumed an asexual, iteroparous organism whose population remains stable during evolution (see Box 3.3). When density dependence acts early during ontogeny, eqn 6.4 also equals the reproductive value at birth (Kozłowski et al. 2004), a measure that many researchers consider an ideal estimate of ﬁtness (Stearns 1992). Their hypothetical organism grows according to an asymptotic function, ﬁrst described by Bertalanffy (1960): s(x) = s∞

(1 − Be−kx ),

(6.5)

where s(x) equals size in length at age x, s∞ equals the asymptotic size, B equals 1−s(0)/s∞ , and k determines the rate of growth between birth and maturation. For our purposes, assume temperature affects growth by changing the value of k. Our hypothetical relationship between temperature and k resembles a typical performance curve (see Chapter 3), such

that a thermal optimum exists for k. For simplicity, we shall restrict our attention to an organism that ceases to grow at maturation; this restriction will not inﬂuence our qualitative predictions about reaction norms (Berrigan and Koella 1994). Berrigan and Koella assumed vital rates depended directly on size or age (Fig. 6.4). Once the organism passes a minimal size (smin ), fecundity increases linearly with an increase in size at maturity: m(α) = f [s(α) − smin ],

(6.6)

where f determines the effect of size on fecundity. Survivorship was calculated by assuming constant rates of mortality for juveniles (µj ) and adults (µa ). Under these strict assumptions, the net reproductive rate equals R0 =

∞ l(x)m(x)dx α

= e−µj α

∞

e−µa (x−α) f s∞ 1 − Be−kx − smin dx. α

(6.7) By differentiating eqn 6.7 with respect to α and setting the resulting equation to zero, Berrigan and Koella found the optimal age at maturity (α ∗ ): fs∞ B(k + µj ) ln µ f (s −s ) ∞ j min . (6.8) α∗ = k The optimal size at maturity was then calculated as follows: ∗

s(α ∗ ) = s∞ (1 − Be−kα )

(6.9)

164

T H E R M A L A D A P TAT I O N

Using eqns 6.8 and 6.9, Berrigan and Koella explored the effects of growth and survivorship on the optimal age and size at maturity. By solving the optimal ages and sizes for a range of parameter values, they constructed optimal reaction norms. In doing so, they implicitly assumed plasticity imposes no cost. Instead of assuming a ﬁxed growth trajectory, other modelers have determined the optimal age and size at maturity for organisms that allocate energy freely between growth and reproduction (reviewed by Kozłowski 1992; Kozłowski et al. 2004; Perrin and Sibly 1993). Although organisms also use energy for maintenance and activity, most researchers have determined how organisms should use energy for production after meeting other demands. Temperature inﬂuences the optimal life history by constraining the rate of assimilation and the cost of maintenance. Prior to maturation, an organism allocates 100% of its surplus energy to growth. The diversion of some energy to reproduction marks the onset of maturation; however, body size can continue to increase after maturation, producing a pattern referred to as indeterminate growth. For comparison with the model of Berrigan and Koella, we shall focus on a simple model of optimal energy allocation described by Kozłowski and colleagues (2004). In this model, growth and fecundity are limited by the organism’s rate of production (or P, since this is a measure of performance). Production depends on the rates of anabolism (A) and catabolism (C), measured in units of energy per time: P = A(M) − C(M) = aMc − bMd ,

(6.10)

where M equals body size in units of energy, a and b determine rates of anabolism and catabolism, and c and d deﬁne the allometries of anabolism and catabolism. The organism grows and reproduces according to its allocation of surplus energy: dM = p(x)P(M) dx

(6.11)

and m(x) =

(1 − p(x))P(M) , n

(6.12)

where p(x) equals the proportion of energy allocated to growth at age x, and n equals the energy required to produce each offspring. As in Berrigan

and Koella’s model, let us assume the mortality rates of juveniles (µj ) and adults (µa ) remain constant (for optimal strategies under size-speciﬁc mortality, see Kozłowski et al. 2004). An optimal age and size at maturity exist as long as the increase in production rate (dP/dM) slows during growth (Kozłowski et al. 2004; Perrin and Sibly 1993); this condition occurs when the allometry of catabolism exceeds the allometry of anabolism (i.e., d > c). In a constant environment, the optimal strategy is quite simple: allocate all surplus energy to growth (p = 1) up to a certain age, and allocate all surplus energy to reproduction (p = 0) thereafter. But how big should an organism grow before switching to reproduction? An organism reaches its optimal size at maturity when the energetic beneﬁt of delaying maturation equals the survival cost: dP(M) = µj (6.13) dM Using this result, Kozłowski and colleagues could solve for the optimal age and size at maturity over a range of environmental conditions. By linking together the optimal solutions, they arrived at optimal reaction norms for age and size at maturity. Like Berrigan and Koella, Kozłowski and colleagues assumed plasticity imposes no cost.

6.3.2 Thermal effects on juvenile mortality Using the models described above, let us consider the optimal reaction norms for three scenarios: (1) juvenile mortality remains constant as the growth rate increases, (2) juvenile mortality decreases as the growth rate increases, and (3) juvenile mortality increases as the growth rate increases. Each of these scenarios corresponds to particular assumptions about the thermal performance curves for survivorship and growth (Fig. 6.5). Which of these scenarios (if any) might explain the temperature– size rule? For Berrigan and Koella’s model, we answer this question by making juvenile mortality (µj ) a function of the growth parameter (k). When juvenile mortality remains constant, an increase in the growth parameter favors earlier maturation at a larger size (Fig. 6.6a), which differs from the pattern observed in most ectotherms (see Section 6.2). When juvenile mortality decreases as the growth

S: 0 G: +

Survivorship

S: – G: –

(b)

S: + G: +

S: – G: +

Survivorship

Growth

Temperature

S: – G: –

(c)

Growth

Performance

S: + G: +

Performance

(a)

Performance

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

S: + G: +

165

S: – G: –

Survivorship

Growth

Temperature

Temperature

Figure 6.5 Hypothetical thermal sensitivities of survivorship (S) and growth (G) depict certain assumptions in the context of thermal performance curves. Survivorship can either increase (+) or decrease (−) with increasing growth rate. The dashed vertical lines mark temperatures at which the covariation between survivorship and growth changes qualitatively.

parameter increases, the prediction does not differ qualitatively from the ﬁrst scenario; in fact, an increase in both growth rate and survivorship should lead to an even larger size at maturity than would an increase in growth rate alone (Fig. 6.6b). Either scenario 1 or 2 might explain exceptions to the temperature–size rule, but scenario 3 predicts a pattern that follows this rule. When an increase in the growth parameter causes a substantial increase in juvenile mortality, an organism should mature earlier at a smaller size (Fig. 6.6c). Relaxing assumptions about growth and fecundity does not alter the main conclusion: higher temperatures must increase juvenile mortality for the optimal reaction norm to follow the temperature–size rule (Sibly and Atkinson 1994). The model by Kozłowski and colleagues can also explain the temperature–size rule when high temperatures simultaneously raise rates of production and mortality (Fig. 6.7). For both models, the cost of greater mortality at higher temperatures must overwhelm the beneﬁt of fecundity associated with rapid growth; otherwise, higher temperatures should lead to delayed maturation at a larger size. Although the two models make similar predictions, the hypothetical mechanism differs greatly (Fig. 6.8). In Berrigan and Koella’s model, the growth of an organism must decelerate with age, even though production clearly accelerates with age (see Fig. 6.4). In Kozłowski et al.’s model, a change in the capacity for production with age favors the diversion of energy from growth to reproduction. Despite very different assumptions about energetics, both models identify a plausible explanation

for the temperature–size rule: higher temperatures reduce the survivorship of juveniles. A similar result emerges when higher temperatures limit the lifespan of an organism (Kozłowski et al. 2004; Perrin and Rubin 1990). Still, do higher temperatures actually decrease the survivorship of real organisms? And, if so, is the thermal sensitivity of survivorship sufﬁcient to explain the temperature–size rule? For a simple model of optimal allocation, my colleagues and I calculated the minimal thermal sensitivity of survivorship needed to explain the temperature–size rule (Angilletta et al. 2004a). This thermal sensitivity depends on two aspects of an organism’s energetics: (1) the thermal sensitivities of anabolism and catabolism (Q10 s of a and b, respectively) and (2) the allometries of anabolism and catabolism (c and d, respectively). For a realistic range of parameters, we asked whether observed thermal sensitivities of survivorship exceed the minimal values required by the model. Two kinds of data bear on this issue. Data from artiﬁcial environments include physiological sources of mortality but exclude ecological sources (e.g., predation). Data from natural environments include both sources, but mortality results from processes occurring at ﬂuctuating temperatures. Based on more than 100 experimental studies in artiﬁcial environments, we concluded temperature has no general physiological effect on survivorship. Even when a higher temperature decreased survivorship, the effect was well below that required by the model (Fig. 6.9). Comparative studies in natural environments generally

166

T H E R M A L A D A P TAT I O N

(b) 1.00

(a) 1.00

mj = 0.004

0.75

0.75

mj = 0.1

0.5

Size

Size

mj = 0.02 mj = 0.5

0.25

0.5

0.25

0

0 0

50

100

150

200

0

50

Age

(c)

100 Age

150

200

1.00 2 3

Size

0.75 1 0.5

0.25

0 0

50

100 Age

150

200

Figure 6.6 Optimal reaction norms for age and size at maturity depend on the covariation between growth rate and juvenile mortality. These plots show optimal reaction norms for three patterns of covariation: (a) juvenile mortality remains constant as growth rate increases, (b) juvenile mortality decreases as growth rate increases, and (c) juvenile mortality increases as growth rate increases. In each plot, the dotted lines depict growth curves and the solid line deﬁnes the corresponding optimal reaction norms for age and size at maturity in response to variation in growth rate. The dashed line marks the minimal size at maturity. In (a), reaction norms for several rates of juvenile mortality (µj ) are shown. In (c), optimal reaction norms for three different relationships between growth rate and juvenile mortality are shown (1–3). The optimal reaction norm follows the temperature–size rule only when juvenile mortality increases appreciably with increasing growth rate (c, line 3). Adapted from Berrigan and Koellas (1994) with permission from Blackwell Publishing.

support the idea that higher temperatures decrease the survivorship of animals; however, the thermal sensitivities still fell short of the critical values (reviewed by Angilletta et al. 2004a). Taken together, these data suggest a complete explanation for the temperature–size rule will require at least one additional mechanism.

6.3.3 Thermal constraints on maximal body size Physiological constraints can also generate optimal reaction norms consistent with the temperature– size rule (Atkinson 1996a; Berrigan and Charnov 1994; Kindlmann et al. 2001). Although warm

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

P = aM 0.67 – bM 0.75

0.12

Temperature

0.08

b

a

25

0.010

0.00200

20 15

0.009 0.008

0.00225 0.00250

10

0.007

0.00275

5

0.006

0.00300 0.0009 Mortality rate

Production rate

167

0.04

0.0006

0.0003 0

0

0

10

20

30

40

50 60 Body size

70

5

10 15 20 25 30 Temperature

80

90

100

Figure 6.7 The optimal reaction norm might follow the temperature–size rule if high temperatures simultaneously raise rates of production and mortality. To illustrate this point, optimal reaction norms are shown for three different thermal sensitivities of mortality rate. The lines connecting the ﬁve production curves represent optimal reaction norms for size at maturity. In all cases, production occurs more rapidly at higher temperatures; parameter values used to calculate production rate (P ) are listed in the upper left corner. Reproduced from Kozłowski et al. (2004) with permission from Oxford University Press.

Body length

(a)

Anabolism Catabolism

(b)

Growth Reproduction Age Figure 6.8 Models of optimal age and size at maturity differ fundamentally in their assumptions about energetics. (a) Berrigan and Koella (1994) assumed an organism allocates an increasing fraction of its energy to catabolism as it ages (represented by the pie diagrams). Consequently, growth must cease at some age because catabolism equals anabolism. (b) In contrast, Kozłowski and colleagues (2004) assumed an organism allocates an increasing fraction of its energy to reproduction as it ages. Consequently, growth ceases even though production continues to increase with age. Reproduced from Czarnołeski and Kozłowski (1998) with permission from Blackwell Publishing.

environments favor rapid growth early in life, they might limit growth later in life. Such constraints could arise when an organism must absorb some limiting resource, such as oxygen, through its body surface. When temperature constrains the maximal size of an organism, the beneﬁt of delayed maturation disappears. When high temperatures constrain the maximal size more severely than low temperatures do, the optimal reaction norm conforms to the temperature–size rule. In a model of energy allocation, we can impose thermal constraints on maximal size by making rates of anabolism and catabolism depend on temperature: P = a(T)Mc(T) − b(T)Md(T) .

(6.14)

The optimal life history follows the temperature size-rule under two conditions. First, temperature can affect catabolism more than it affects anabolism; speciﬁcally, the Q10 of b must exceed the Q10 of a (Perrin 1995). Second, temperature can affect the allometries of anabolism and catabolism in opposite ways; higher temperatures must decrease c while increasing d (Strong and Daborn 1980). In fact, both conditions can work together to favor smaller sizes at higher temperatures. Do these hypothetical

168

T H E R M A L A D A P TAT I O N

2 3 3, 4,

2 3 2, 3,

(a) 0

2 3 4, 5,

2 3 5, 6,

–0.02 –0.04

ln(survivorship at low temperature)

–0.06 –0.08 –0.10 (b)

0

5

10

15

20

0

–0.02 2, 1 3, 2 4, 3

–0.04 –0.06

3, 1 4, 2 5, 3

–0.08 –0.10

0

5 10 15 Q10 ln(survivorship)

20

Figure 6.9 The thermal sensitivity of juvenile survivorship needed to explain the temperature–size rule depends on the rates of anabolism and catabolism. (a) The case in which the allometries of anabolism and catabolism are identical (i.e., c = d ). (b) The case in which anabolism scales allometrically with body mass (c < 1) and catabolism scales isometrically (d = 1). The lines are isoclines for realistic combinations of thermal sensitivities of anabolism and catabolism, which are listed in the margin as [Q10 of anabolism, Q10 of catabolism]. To the right of each isocline, the optimal size at maturity decreases with increasing temperature in accord with the temperature–size rule. The points are thermal sensitivities of survivorship for 130 populations of 114 species of ectotherms. Reproduced from Angilletta et al. (2004a) with permission from Oxford University Press.

constraints on maximal size help to explain patterns of age and size at maturity? Empirical evidence strongly refutes the hypothesis that temperature speeds catabolism more than anabolism. If this condition applies, organisms must grow less efﬁciently in warmer environments

(Box 6.2). This prediction can be widely tested because physiologists have measured the thermal sensitivity of growth efﬁciency for many species of arthropods, mollusks, and ﬁsh. In a survey of more than 50 species, Angilletta and Dunham (2003) concluded that ectotherms usually grow more rapidly and more efﬁciently in warmer environments (Fig. 6.10). Even when data for a particular species do support this hypothesis, the range of temperatures over which this mechanism can explain the temperature–size rule rarely exceeds 5◦ C (but see Karl and Fischer 2008). Therefore, this explanation applies only to a narrow range of temperatures in the minority of ectotherms studied to date. In contrast, available evidence strongly supports the hypothesis that allometries of anabolism and catabolism respond differently to temperature. In a study of isopods (Idotea baltica), ingestion scaled almost isometrically (c = 0.94) at a low temperature but allometrically (c = 0.71) at a high temperature, whereas respiration scaled allometrically (d = 0.68) at a low temperature and isometrically (d = 1.00) at a high temperature. Consequently, growth decelerated with age in the warm environment but accelerated with age in the cold environment. Similar trends have been documented for several other species (see Angilletta and Dunham 2003). Still, the true generality and physiological basis of this phenomenon remain unknown. To predict the temperature–size rule, these models require production to decelerate as an organism ages. Kindlmann and colleagues (2001) assumed that production accelerated early in ontogeny but decelerated late in ontogeny because of senescence. Under these conditions, the growth of juveniles accelerates with age while the fecundity of adults decelerates with age. This senescence of production can result from a decrease in resource acquisition or an increase in required maintenance. The optimal age and size at maturity depends on the rate of senescence; if senescence proceeds more rapidly at higher temperatures, early maturation enables reproduction before senescence takes a major toll on fecundity. Kindlmann and colleagues claimed their model likely explains the general effect of temperature on body size. However, their conclusion depends on the validity of the Malthusian parameter

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

169

Box 6.2 Growth efﬁciency and the temperature-size rule

K2 =

aM 2/3 − bM 1 A−C = . A aM 2/3

(6.15)

To explore the inﬂuence of body temperature on the net growth efﬁciency, we made a and b functions of temperature, as assumed by Perrin (1995): K2 =

a(Tb )M 2/3 − b(Tb )M 1 , a(Tb )M 2/3

(6.16)

as a criterion of ﬁtness, the time-course of senescence, and the thermal sensitivities of assimilation, gonadogenesis, and senescence. As such, we cannot infer the generality of this hypothesis at the present time.

6.3.4 Thermal effects on population growth A positive effect of temperature on the growth of a population can also favor the temperature–size rule. When density dependence prevents the growth of a population, the timing of reproduction has no direct bearing on the ﬁtness of an organism. In other words, two offspring produced late in life beneﬁt an organism as much as two offspring produced early in life. In a growing population, the situation differs greatly; all else being equal, an organism that reproduces early in life enjoys greater ﬁtness than one that reproduces late in life (see Box 3.3). Therefore, early maturation at a small size might reﬂect a greater potential for population growth in warm environments (Sibly and Atkinson 1994). In laboratory experiments, the Malthusian parameter depends on temperature in much the same way that other performances do (Huey and

which simpliﬁes to K2 = 1 −

b(Tb ) 1/3 M . a(Tb )

(6.17)

Taking the derivative of K2 with respect to body temperature yields the following:

b (Tb )a(Tb ) − b(Tb )a (Tb ) dK2 . (6.18) = −M 1/3 dTb a(Tb )2 Note the derivative of K2 with respect to body temperature is negative when the following condition is met: a(Tb ) a (Tb ) < . b (Tb ) b(Tb )

(6.19)

Perrin’s model requires this condition by assuming b (Tb ) exceeds a (Tb ) and a(Tb ) exceeds b(Tb ) for a growing organism. Therefore, K2 must decrease with increasing temperature if Perrin’s model accurately describes growth.

70 Gross (K1)

60 % of observations

Recently, Dunham and I derived an implicit assumption of Perrin’s (1995) explanation for the temperature–size rule: an organism must grow less efﬁciently in a warmer environment (Angilletta and Dunham 2003). We started by redeﬁning Bertalanffy’s equation in terms of growth efﬁciency. Bertalanffy (1960) deﬁned the rates of anabolism and catabolism as aM2/3 and bM1 , respectively. These rates determine what physiologists refer to as the net growth efﬁciency (K2 ):

Net (K2)

50 40 30 20 10 0 Positive

Negative Unimodal Insignificant

Figure 6.10 An increase in temperature usually enhances the growth efﬁciency of an ectotherm. Relationships between temperature and growth efﬁciency (gross or net) were characterized as positive, negative, unimodal, or statistically insigniﬁcant. These gross and net growth efﬁciencies represent 89 populations of 53 species and 24 populations of 20 species, respectively. Data are from Angilletta and Dunham (2003). Reproduced from Angilletta et al. (2004) with permission from Oxford University Press.

Berrigan 2001). Therefore, this hypothesis provides a plausible explanation for the temperature–size rule when similar temperatures maximize the growth of an individual and the growth of its population.

170

T H E R M A L A D A P TAT I O N

Unfortunately, thermal sensitivities of population growth have generally been estimated under artiﬁcial conditions, which excluded many sources of density dependence such as competition and predation. Until we have more information about the thermal sensitivity of population growth in natural environments, we cannot conﬁrm the generality of this hypothesis. Geographic clines might also stem from thermal effects on population growth. Lower latitudes and altitudes offer longer seasons that enable more generations per year. Even if a population remains stable over the long run, natural selection favors early maturation when genotypes can complete multiple generations between periods of density dependence. This situation exists in seasonal environments, where reproduction occurs during part of the year and density-dependent mortality culls the population during the remainder of the year. In such environments, the optimal life history might conform to the temperature–size rule. Some support for this hypothesis exists among populations of butterﬂies (Lycaena hippothoe). Genotypes complete either one or two generations per year depending on their environment. Genotypes from a population that completes two generations per year responded more strongly to developmental temperature than those from populations that complete one generation per year (Fischer and Fiedler 2002). Do not confuse this variation in thermal plasticity among populations

with the variation in body size associated with multivoltinism in insects. In the latter case, genotypes evolve small size at maturity in environments where they can achieve two or more generations per year (Masaki 1967), but the phenomenon does not require plasticity.

6.3.5 A synergy of evolutionary mechanisms In summary, any one of three mechanisms can cause the evolution of reaction norms consistent with the temperature–size rule: thermal constraints on juvenile mortality, maximal size, and population growth (Table 6.1). Support for each of these hypotheses varies greatly among species, even though the phenomenon we seek to explain seems nearly universal. Before we abandon hope of making sense of this situation, let us brieﬂy consider the potential synergism among these hypothetical mechanisms. The overwhelming generality of the temperature– size rule has had an unfortunate effect on the way many researchers have sought to understand thermal reaction norms. Atkinson (1996a) recognized this effect when he stated the following: The discovery of a widespread relationship such as a temperature–size rule tends to direct research towards general explanations which apply throughout ectotherms, and away from those speciﬁc to particular populations, species or groups of species. Simple, rather than complex explanations are usually sought.

Table 6.1 Hypothetical mechanisms that can generate optimal reaction norms consistent with the temperature–size rule. These mechanisms are identiﬁed as extrinsic or intrinsic factors Hypothetical mechanism

Predicted relationship

Source

Extrinsic factors Juvenile mortality increases with temperature

Negative

Adult mortality decreases with temperature Rate of population growth increases with temperature Seasonal constraints on growth and development

Negative Negative Saw-toothed

Kozłowski et al. (2004); Sibly and Atkinson (1994); Berrigan and Koella (1994) Charnov and Gillooly (2004); Sibly and Atkinson (1994) Sibly and Atkinson (1994) Roff (1980)

Intrinsic factors Growth efﬁciency decreases with temperature Lifespan decreases with temperature Rate of gonadogenesis increases with temperature

Negative Dome-shaped Negative

Atkinson (1996b); Berrigan and Charnov (1994) Perrin (1988) Kindlmann et al. (2001)

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

6.4 General patterns of reproductive allocation 6.4.1 Thermal plasticity of offspring size Temperature appears to exert a general effect on reproductive allocation as well as age and size at maturity. Atkinson and colleagues (2001) reviewed 36 experiments in which researchers measured the effects of temperature on offspring size (or egg size). More than 70% of these experiments revealed a decrease in offspring size with increasing temperature; in fact, only 3% reported the opposite relationship (Fig. 6.11). The same trend was observed for relative offspring size (offspring size/maternal size), suggesting the pattern was not merely a consequence of the temperature–size rule. More recent experiments yielded additional support for this conclusion; lower temperatures caused females to produce larger offspring (Fischer et al. 2003a; Karlsson and Van Dyck 2005; Stelzer 2002) or caused no clear change in offspring size (Du et al. 2005; Tveiten et al. 2001), but did not lead to smaller offspring.

30 25 Number of cases

On the one hand, simple explanations appeal intuitively to our human brains (despite how the world repeatedly demonstrates its complexity). On the other hand, simple explanations usually fall very short of offering a meaningful understanding of broad patterns of life-history variation (Angilletta and Dunham 2003; Angilletta et al. 2004a). Rather than accept any of these simple explanations, we should expect a multitude of factors to drive the evolution of thermal reaction norms. Indeed, all three of the mechanisms discussed in this section could operate in synergy. Sibly and Atkinson (1994) noted the conditions for the evolution of large adults in cold environments were more likely to occur when temperature affected both juvenile mortality and population growth. Likewise, Kozłowski and colleagues (2004) modeled a synergism between the thermal sensitivities of juvenile mortality and maximal size. Although complex explanations often fail to win the hearts of biologists, the points made by these researchers should not be ignored. Ultimately, we will need to integrate our knowledge of physiology and ecology to arrive at a general explanation for the thermal plasticity of age and size at maturity.

171

Raw size Relative size

20 15 10 5 0 Negative No effect Positive Thermal effect of offspring size

Figure 6.11 In ectotherms, mothers experiencing higher temperatures generally produce smaller offspring. Only a minority of populations studied in the laboratory consisted of genotypes that deviated from this pattern. This conclusion also held when the effect of maternal size was removed by computing a relative offspring size (offspring size/maternal size). Data were summarized by Atkinson et al. (2001).

For some species, we know females can adjust their egg size very rapidly in response to a change in temperature. For example, the yellow dung ﬂy (Scathophaga stercoraria) typically produces large eggs when developing at a low temperature in natural or artiﬁcial environments (Blanckenhorn 2000). To determine whether females could respond rapidly to environmental temperature, Blanckenhorn (2000) took females that had developed at high temperatures and shifted them to a low temperature after they had produced their ﬁrst clutch. As he expected, females produced larger eggs after the switch to a lower temperature. The phenotypic plasticity that Blanckenhorn observed in the laboratory mirrored a seasonal change he had documented in the natural environment. At least for dung ﬂies, thermal plasticity of egg size can enhance ﬁtness during thermal changes within generations.

6.4.2 Thermal clines in offspring size Offspring size (or egg size) also varies greatly along geographic clines. Much of this variation accords with the patterns of thermal plasticity observed in controlled environments. Thermal clines in egg size have been documented among arthropods (Fischer and Fiedler 2001b; Fox and Czesak 2000;

172

T H E R M A L A D A P TAT I O N

Parry et al. 2001; Wilhelm and Schindler 2000), amphibians (Morrison and Hero 2003b), and ﬁsh (Laptikhovsky 2006). In most cases, individuals in colder environments produced larger eggs (but see Fleming and Gross 1990). In D. melanogaster, thermal clines evolved independently in Australia and South America (Azevedo et al. 1996). An interesting “experiment” in nature supported temperature as the cause of local variation in egg size. Meffe (1990) compared eggs of mosquitoﬁsh (Gambusia holbrooki) in a natural pond with those of ﬁsh in a polluted pond; the pollution consisted of thermal efﬂuent, which periodically raised the temperature of the pond to as high as 45◦ C. Fish in the polluted pond produced smaller offspring throughout the year than did ﬁsh in the unpolluted pond. If we consider these patterns in the context of thermal plasticity, we might conclude that a model of thermal adaptation could explain geographic variation in egg size. Although thermal clines in egg size resemble thermal reaction norms, not all geographic variation can be attributed to plasticity. For a few species, researchers have documented the contributions of genetic and environmental factors by raising populations in the laboratory or by transplanting populations to new environments. For example, Armbruster and colleagues (2001) raised pitcherplant mosquitoes from six populations in a common environment. All mosquitoes experienced ideal summer conditions, yet mosquitoes from higher latitudes laid larger eggs. The variation in egg size among populations diminished when mosquitoes were raised at high densities, suggesting intraspeciﬁc competition plays some role in the reproductive decision. Given the common environment, patterns at high and low densities likely reﬂect the genetic divergence of egg size among populations; unfortunately, patterns of thermal plasticity within populations were not quantiﬁed. Parallel clines in the egg size of D. melanogaster also have a genetic basis; ﬂies from higher latitudes laid larger eggs, even after ﬂies from all populations were raised at constant temperatures for more than a year (Azevedo et al. 1996). In contrast, a transplant experiment revealed plasticity as the major cause of a thermal cline in the land snail, Arianta arbustorum (Baur and Raboud 1988). This species occurs at altitudes up to 2700 m, where the relative egg size exceeds that observed at lower

altitudes. When snails were transplanted from ﬁve sites between 1200 and 2600 m to a site at 500 m, the majority of snails laid eggs similar in size to those laid in their natal environment. However, snails transplanted from the highest altitude laid smaller eggs than did snails that remained at this altitude. In this case, the thermal cline in egg size resulted from the combined effects of genetic and environmental factors. These experiments highlight the need to identify the sources of variation in offspring size before devising theories to explain thermal clines.

6.4.3 Experimental evolution of offspring size Three types of experimental evidence prompt an adaptive interpretation of the thermal plasticity of offspring size: (1) selection gradients at different temperatures, (2) additive genetic variances of reaction norms, and (3) responses to selection in controlled environments. Only two experiments have been conducted to identify the selective beneﬁts of egg size over a range of temperatures. Blanckenhorn (2000) raised dung ﬂies (Scathophaga stercoraria) at 19◦ C and divided mature siblings among four thermal environments (11◦ C, 15◦ C, 19◦ C, and 23◦ C). Eggs produced at each temperature were then divided among the same thermal environments to document variation in rates of mortality, development, and growth. Surprisingly, both large eggs produced at 11◦ C and small eggs produced at 23◦ C survived better than the intermediate eggs produced at 15◦ C or 19◦ C. Moreover, hatchlings from eggs produced at 23◦ C survived the best at all temperatures except 23◦ C! Rates of development and growth by offspring depended primarily on their thermal environment. Since temperature had little effect on fecundity, the patterns uncovered by Blanckenhorn’s experiment seem very puzzling. A similar experiment by Fischer and colleagues (2003b) focused on the selective advantage of egg size for butterﬂies in cold and hot environments. Females of Bicyclus anynana alter the size of their eggs rapidly in response to environmental temperature (Fig. 6.12). After raising families of butterﬂies at 23◦ C, these researchers divided sisters between 20◦ C and 27◦ C to obtain eggs of different sizes. Then, they reciprocally transferred eggs

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

Egg size (mm2)

0.72

0.68

0.64

0.60 1

2 Measurement

Figure 6.12 Egg size of the butterﬂy Bicyclus anynana responded rapidly to changes in environmental temperature. Females were raised at one temperature and either were maintained at this temperature as adults or were switched to a different temperature. Measurements of egg size occurred just after development at the ﬁrst temperature (1) and 10 days after being switched to the second temperature (2). Symbols are as follows: circles, 20–20◦ C; squares, 20–27◦ C; triangles, 27–20◦ C; and diamonds 27–27◦ C. Adapted from Fischer et al. (2003a) with permission from the Royal Society.

produced at each temperature to see whether the selective advantage of egg size depended on temperature. Females produced small eggs at 27◦ C, but these eggs suffered relatively high rates of mortality at 20◦ C. Consequently, females that laid fewer, larger eggs would expect more of their offspring to reach maturation at 20◦ C. Yet the beneﬁt of fecundity associated with small eggs outweighed the cost of survivorship at 27◦ C. In a subsequent experiment, these researchers found that eggs produced at 20◦ C and 27◦ C suffered equally during starvation and desiccation (Fischer et al. 2003c). Thus, the mechanism by which large size enhances the survivorship of offspring in cold environments remains elusive. Using a half-sibling design, Steigenga and colleagues (2005) estimated the additive genetic variance of thermal reaction norms for egg size in B. anynana. Genetic correlations between environmental temperatures were positive but signiﬁcantly less than 1.0, indicating a potential for thermal plasticity to respond to natural selection. Unfortunately, the experimental design confounded environmental temperature and laying date, creating doubt about

the true magnitude of thermal plasticity. Interestingly, parent–offspring and full sibling analyses of the same species did not detect genetic variation in reaction norms (Fischer et al. 2004), despite clear plasticity of egg size in response to temperature. If we assume no effects of dominance and epistasis, maternal effects on egg size far exceeded genetic effects (see Box 1.2). Azevedo and colleagues (1996) exposed replicated lines of D. melanogaster to constant temperatures. Originally, selection lines were initiated at 16.5◦ C and 25◦ C, but an additional set of lines were established at 29◦ C by splitting the lines at 25◦ C. After 9 years of selection at either 16.5◦ C or 25◦ C and 4 years of selection at 29◦ C, ﬂies from all lines were raised at 25◦ C and their eggs were compared. Flies selected at 16.5◦ C laid larger eggs than ﬂies selected at either 25◦ C or 29◦ C, but ﬂies from these warmer selective environments laid eggs of similar sizes. These researchers also compared the thermal reaction norms of ﬂies evolving at 16.5◦ C and 25◦ C. Females from these selective environments were raised at 25◦ C and then oviposited sequentially at 16.5◦ C and 25◦ C. Selection had not altered the thermal plasticity of egg size, but eggs laid by cold-adapted ﬂies were 5–10% larger than eggs laid by warm-adapted ﬂies (Fig. 6.13). 0.013

Egg volume (mm3)

0.76

173

0.012

0.011

0.010

16.5

25 Temperature (°C)

Figure 6.13 Experimental evolution led to a divergence in mean egg size between populations of Drosophila melanogaster, but the thermal plasticity of egg size remained similar among selection lines. Dashed and solid lines represent reaction norms for populations evolving at 16.5◦ C and 25◦ C, respectively. Adapted from Azevedo et al. (1996) with permission from Blackwell Publishing.

174

T H E R M A L A D A P TAT I O N

This body of evidence suggests three points. First, selection favors particular reaction norms for egg size. Second, these reaction norms persist despite heritable variation. Finally, changes in environmental temperature elicit genetic responses in egg size that parallel plastic responses. Given these points, we now turn to models of optimal reproductive allocation to identify possible causes for the thermal plasticity of egg size.

6.5 Optimal size and number of offspring The theory of optimal reproductive allocation attempts to deﬁne the environmental causes of variation in reproductive traits, such as the size and number of offspring. Since the seminal model of Smith and Fretwell (1974), the theory has grown around a set of core assumptions: (1) reproductive decisions evolve to maximize the product of the number of offspring and their probability of surviving; (2) the ﬁtness of offspring increases asymptotically as their share of resources increases; and (3) mothers have a ﬁxed quantity of resources available for reproduction. Under these conditions, a single optimal size should dictate the allocation of resources among offspring (Smith and Fretwell 1974). Subsequent modeling (McGinley et al. 1987) supported the existence of an optimal offspring size, except under rare conditions when a mixture of sizes confers greater ﬁtness (see Section 6.6). In this section, we shall see how thermal heterogeneity affects the optimal size and number of offspring—be they seeds, eggs, or babies.

6.5.1 Direct effect of temperature on the optimal offspring size Changes in temperature over space and time should favor variation in offspring size. Among ectotherms, the performance of offspring decreases as temperatures rise above or fall below the thermal optimum (see Chapter 3). The optimal offspring size depends on several performances of offspring. Speciﬁcally, the optimal size decreases as (1) the growth rate of offspring increases, (2) the mortality rate of offspring decreases, and (3) the growth rate of the population increases (Sibly and Calow 1983; Taylor

and Williams 1984). Because temperature potentially affects all three of these conditions, ectotherms should exhibit thermal plasticity of offspring size (Perrin 1988). To understand the effects of temperature on the optimal offspring size, we can use a simple model described by Yampolsky and Scheiner (1996). Like previous modelers, Yampolsky and Scheiner assumed the survivorship of offspring increased asymptotically with increasing size. Furthermore, the survivorship of offspring continued to increase as they approached maturity. To model this phenomenon, they let the rate of juvenile mortality (j) decrease exponentially as offspring grew in size (s): j(s) = j0 e−ks ,

(6.20)

where j0 is the rate of juvenile mortality when s = 1, and k determines the size dependence of mortality. Importantly, they also assumed the rate of mortality did not depend on temperature. Offspring grew according to a linear function: s(x) = s0 +

sα − s0 x, α

(6.21)

where s0 and sα equal sizes at birth and maturation, respectively. Thus, Yampolsky and Scheiner implicitly assumed that energy acquisition and allocation remained constant during the juvenile stage. With these assumptions in place, they solved for the ﬁtness of a genotype in a growing population (r). This calculation became greatly simpliﬁed when the genotype reproduces semelparously. In this special case, ﬁtness equals r=

ln l(α) + ln m(α) . α

(6.22)

By sequentially substituting eqns 6.20 and 6.21 into eqn 6.22, they arrived at a function relating ﬁtness to several life-history traits, including size at birth:

j0 e−ksα − e−ks0 ln m(α) + (6.23) r= k(sα − s0 ) α From this equation, we can see that r depends strongly on the age at maturity (α) and the rate of juvenile mortality (as inﬂuenced by j0 and k). Yampolsky and Scheiner used eqn 6.23 to infer the selective pressures imposed by changes in temperature. Two effects stand out. First, a genotype that

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

matures at a younger age (i.e., a lower α) would enjoy a greater chance of reaching adulthood and hence would experience a greater r. This effect is reﬂected by an increase in the right-hand term of eqn 6.23 with a decrease in α. Second, a genotype that suffers less mortality (i.e., a lower j0 ) would also experience a greater r. This effect is reﬂected by an increase in the left-hand term of eqn 6.23 with a decrease in j0 (note that e−ksα − e−ks0 < 0). Both effects cause selection for smaller offspring because of the concomitant increase in fecundity. From Fig. 6.14, we can see that selection favors small offspring whenever they would mature at an early age or experience little risk of mortality (see also Taylor and Williams 1984). Because ectotherms tend to grow faster and mature earlier in warmer environments (see Section 6.2), Yampolsky and Scheiner reasoned that individuals in warm environments should produce

small offspring. Unfortunately, we cannot generalize the effect of environmental temperature on juvenile survivorship (see Section 6.3.2). If juveniles suffer greater mortality in warmer environments, mothers should produce relatively large offspring. Thus, opposing selective pressures could result from thermal sensitivities of growth and development and the thermal sensitivity of juvenile survivorship. The predictions of this model depend greatly on the underlying assumption about the relationship between the size and performance of offspring. As with most models of optimal reproductive allocation, this one assumes large offspring always outperform small offspring. The model breaks down when the relationship between size and performance differs among thermal environments, either through acclimation or through interactions between temperature and other factors (see, for

(c) 1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

Offspring size

(a) 1.0

0 101

102

103

0 (d) 1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

Offspring size

(b) 1.0

0

175

101

102

103

101

102 a . j0

103

0 101

102 a . j0

103

Figure 6.14 The optimal offspring size increases as either the age at maturity (α) or the rate of mortality (j0 ) increases. (a–d) Equilibrial reaction norms given for four relationships between offspring size and juvenile mortality (i.e., k increases from plot a to plot d; see eqn 6.20). Both stable equilibria (solid lines, ﬁlled circles) and unstable equilibria (dashed lines, open circles) are shown. Stable equilibria occur only when an increase in offspring size results in a large decrease in juvenile mortality (c and d). Under these conditions, we should expect selection to favor an increase in egg size with decreasing environmental temperature. Adapted from Yampolsky and Scheiner (1996, The American Naturalist, University of Chicago Press). © 1996 by The University of Chicago.

176

T H E R M A L A D A P TAT I O N

example, Gagliano et al. 2007). One can envision potential disadvantages of large size in both aquatic and terrestrial environments. For example, large embryos could fail to acquire sufﬁcient oxygen from warm waters, as oxygen delivery scales disproportionally to body size (Woods 1999). Additionally, large eggs might attract the attention of predators because of their great energetic value (Stephens and Krebs 1986). In reality, we do not know the true relationship between the size and performance of offspring for most species in most environments (but see Sinervo et al. 1992). Therefore, we should cautiously recognize that thermal plasticity of egg size could be caused by something other than thermal sensitivities of growth and development.

6.5.2 Indirect effects of temperature on the optimal offspring size The optimal reproductive strategy depends not only on the thermal sensitivity of offspring performance but also on the thermal sensitivity of the maternal phenotype. Adults in warm environments differ dramatically from adults in cold environments. As discussed in Section 6.2, a mother in a cold environment would likely be older and larger than a mother in a warm environment. These differences would affect the relative capacities for energy acquisition and parental care—capacities that undoubtedly inﬂuence reproductive decisions. Large mothers would also be more capable of passing large offspring during birth or oviposition. Therefore, the thermal plasticity of age and size at maturity mediates indirect effects on the optimal reproductive decision. The capacities of large mothers to acquire and store more energy greatly affect the optimal size and number of offspring. Parker and Begon (1986) modeled a capital breeder that acquired energy at a constant rate until reaching a maximal level. They assumed a large female acquired energy more rapidly and could store energy more readily than a small female. Should a large female use her extra energy to produce more offspring, larger offspring, or both? The answer depends on the relationship between offspring size and ﬁtness. If we assume

the survivorship of offspring increases asymptotically with size (as in most models), all mothers should produce offspring of the same size. A large female should use her additional energy to produce more offspring because the beneﬁt of greater fecundity exceeds the beneﬁt of producing offspring beyond a certain size. The situation changes when the survivorship of offspring depends on more than just their size. Imagine that offspring compete with one another, such that an increase in the number of offspring heightens the intensity of competition. Then a large female that produced many offspring would have her increase in fecundity offset by a decrease in her offspring’s performance. Under this condition, a large female should use her additional energy to produce larger offspring. When the survivorship of offspring depends on both their size and their density, larger mothers should adopt a mixed strategy, producing both more and larger offspring. The care provided by a mother also affects the optimal size of her offspring. Conceivably, larger mothers construct better nests, deter more predators, or provide more food for their offspring. Sargent and colleagues (1987) modeled the effect of parental care on the optimal offspring size. These researchers separated the survivorship of offspring into two stages: embryonic and juvenile stages. Assume that embryos inside larger eggs (or seeds) take longer to develop or suffer greater rates of mortality. In either case, the probability of an offspring reaching the juvenile stage (and ultimately adulthood) decreases with increasing size. In contrast, the rate of juvenile mortality decreases with increasing size. Under these conditions, a mother should produce larger eggs if her care can improve the survivorship of her offspring during their embryonic stage. Parental care offsets the decrement in embryonic survivorship associated with large eggs, such that mothers reap the beneﬁts of providing offspring with the energy needed to become large juveniles. Thus, a large mother should produce fewer, larger eggs than a small mother if she provides better parental care. Hendry and colleagues (2001) predicted the same phenomenon from a very different relationship between the size and survivorship of offspring. Speciﬁcally, they assumed that offspring of an intermediate size enjoyed the greatest

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

survivorship. Yet through some form of parental care, a large mother could ameliorate the disadvantages of extremely large offspring. The two models essentially portray the same idea despite containing different details. Finally, large mothers escape the constraints on offspring size imposed by morphology. Most animals pass their eggs or babies through apertures in their internal or external skeletons. The maximal size of the aperture determines the maximal size of the offspring. Consequently, small mothers can be forced to produce smaller offspring than would be optimal for larger mothers (Congdon and Gibbons 1987). This nonadaptive mechanism would generate a unique pattern of variation: offspring size would increase with increasing maternal size up to a maximum that approximates the optimal size in the absence of a morphological constraint (Ouﬁero et al. 2007).

6.5.3 Teasing apart direct and indirect effects on reproductive allocation From the models described in the preceding section, we can safely conclude that temperature should have direct and indirect effects on reproductive decisions (Table 6.2). Direct effects result from the thermal sensitivities of offspring performances. Indirect effects are mediated by maternal size and its consequences for energy acquisition, parental care, and morphological constraints. Because two or more of

177

these effects can occur simultaneously, we should try to distinguish their relative contribution to variation in reproductive allocation. A strong inference approach can tease apart the direct and indirect effects of environmental temperature on the evolution of reproductive traits (Angilletta et al. 2006b). We can use the predictions of optimality models to construct statistical models relating environmental temperature, maternal size, and reproductive traits. A simple path model accounts for all of the hypothetical mechanisms linking environmental temperature to reproductive traits (Fig. 6.15a). This statistical model can be evaluated with phylogenetic comparative data, which describe the correlated evolution of two or more traits (e.g., independent contrasts). Because any or all of these mechanisms could operate simultaneously, we can use Akaike’s information criterion (see Box 3.1) to select the model that best ﬁts the available data. Support for a particular statistical model implies support for its corresponding optimality model(s). Angilletta and colleagues (2006) used this approach to infer the likely causes of reproductive decisions in the eastern fence lizard (Sceloporus undulatus). We found considerable support for the hypothesis derived from Parker and Begon’s model (1986), in which (1) the size of a mother enhances her energy acquisition and (2) the survivorship of offspring depends on their size and their density (Fig. 6.15b). Other statistical approaches can help distinguish between direct and indirect effects. For

Table 6.2 Models of optimal reproductive allocation predict speciﬁc relationships between temperature and reproductive traits under sets of hypothetical conditions (reproduced from Angilletta et al. 2006b)

Hypothetical conditions 1. Offspring grow faster in warmer environments 2. (a) Mothers are smaller in warmer environments, (b) smaller mothers acquire less energy, and (c) offspring survival increases as egg size increases 3. (a) Mothers are smaller in warmer environments, (b) smaller mothers acquire less energy, and (c) offspring survival decreases as density increases 4. (a) Mothers are smaller in warmer environments and (b) offspring survival increases as maternal size increases

Thermal effect on egg size

Thermal effect on clutch size

Source

Decrease None

Increase Decrease

Perrin (1988); Yampolsky and Scheiner (1996) Parker and Begon (1986)

Decrease

None

Parker and Begon (1986)

Decrease

Increase

Hendry et al. (2001); Sargent et al. (1987)

178

T H E R M A L A D A P TAT I O N

(a)

Maternal Size

+2

Clutch size +3/+4

Temperature

(b)

1 4 – /–

Egg size

–1 0.66

Maternal size –0.58 ** Temperature

0.53 0.69 ***

Clutch size

0. 52 * Egg size

1.00 0.73 Figure 6.15 An analysis of the direct and indirect effects of environmental temperature on reproductive traits can reveal how multiple factors simultaneously affect reproductive allocation. (a) A path model depicts the predicted relationships among environmental temperature, maternal size, and reproductive traits under different conditions. These relationships were derived from the optimality models summarized in Table 6.2. The predicted relationships (positive or negative) are linked to speciﬁc conditions by superscripts: 1, Thermal sensitivity of growth; 2, survival depends on size; 3, survival depends on density; 4, survival depends on maternal size. (b) A submodel best described the phenotypic covariation among populations of the eastern fence lizard (Sceloporus undulatus). Data were independent contrasts generated from an intraspeciﬁc phylogenetic hypothesis. Adapted from Angilletta et al. (2006b, The American Naturalist, University of Chicago Press). © 2006 by The University of Chicago.

example, Perrin (1988) used an analysis of covariance to partition the effects of maternal size and environmental temperature on the egg size of Simocephalus vetulus; his analysis clearly showed that the thermal effect on egg size exceeded the maternal effect.

6.6 Optimal variation in offspring size Thermal heterogeneity can also select for females that produce a mixture of offspring sizes. The models described in the preceding section assume that a mother must produce offspring of one size at each reproductive event. Now consider the possibility that a mother could produce two or more sizes of offspring at one time. Under what thermal conditions

would this strategy outperform a single offspring size? We can use a model by McGinley and colleagues (1987) to infer the conditions that favor a mixture of offspring sizes. These researchers modeled the production of offspring in an environment that varied either spatially or temporally. To simplify their analysis, they assumed the environment consisted of good and bad patches, instead of a continuum of patch qualities. For our purpose, good and bad patches would provide optimal and suboptimal temperatures, respectively. Offspring can disperse randomly among patches or they can be directed to particular patches by their parents. Given our previous considerations, we should expect good patches to favor small offspring and bad patches to favor large offspring. But should mothers ever produce a combination of large and small offspring? According to the model, most conditions favor a single offspring size over a mixture of sizes. The optimal offspring size depends on the dispersal of offspring, the frequency of patches, and the degree of density dependence. If offspring disperse randomly among patches, a single offspring size always returns greater ﬁtness than does a mixture of two sizes. Selection for small offspring would occur when patches differ greatly in temperature or when the environment consists mainly of good patches. Otherwise, a mother should produce offspring of the minimal viable size in a bad patch. If a mother can direct her offspring to a good patch, she should produce small offspring; however, anything less than a perfect direction of offspring to good patches favors larger offspring. Thus, spatial heterogeneity in temperature should alter the mean offspring size without adding to its variance. Extreme variation in temperature among generations can select for variation in offspring size (McGinley et al. 1987). A mixture of offspring sizes enhances fecundity while guarding against extinction during generations of extreme temperatures. A mixed strategy should evolve when temperatures differ greatly between good and bad generations but favorable temperatures occur frequently. Otherwise, mothers should produce offspring large enough to survive in a poor thermal environment. Not surprisingly, the condition favoring a mixture of offspring sizes also favors a thermal generalist

T E M P E R AT U R E A N D T H E L I F E H I S TO RY

over a thermal specialist (see Chapter 3). Therefore, selection for variation in offspring size likely occurs in short-lived organisms that have multiple generations per year, because such organisms would experience the greatest variation in temperature among generations.

6.7 Conclusions Optimality models have provided a wealth of competing hypotheses for the thermal plasticities of life-history traits. Reaction norms for the age and size at maturity should evolve according to the thermal sensitivities of juvenile survivorship, maximal size, and population growth. Reaction norms for the size and number of offspring should evolve according to the thermal sensitivity of juvenile performance and the thermal plasticity of maternal size. Although this theory provides potential explanations for general patterns of life-history variation, the critical assumptions of most models remain to be validated. Existing data cast doubt on the assumptions required to explain the temperature–size rule, such as strong thermal effects on juvenile survivorship or maximal size. Insufﬁcient data exist to verify assumptions pertaining to optimal offspring size, such as the relationships among the size, density, and ﬁtness of offspring. More importantly, all of the models implicitly assume that the problems of optimizing adult size and offspring size can be solved independently. Yet, we know that environmental temperature inﬂuences the optimal offspring size indirectly via changes in maternal size. Stelzer (2002) turned this observation around by arguing that thermal effects on egg size contribute to temperature–size relationships. By independently manipulating the temperatures of reproducing rotifers and their growing offspring, he found that the thermal plasticity of egg size contributed substantially to variation in size at maturity. When modeling the optimal size at maturity, researchers have assumed temperature does not affect the size at birth. The fact that temperature does affect the size at birth prompts an important question. Does evolution by natural selection produce a coadapted suite of life-history phenotypes?

179

Comparative and experimental studies of D. melanogaster suggest several life-history traits coevolve in response to temperature. In South America and Australia, ﬂies from higher latitudes developed faster and grew more efﬁciently as larvae (James and Partridge 1995; Robinson and Partridge 2001; van’t Land et al. 1999), reached larger sizes at maturity (James et al. 1995, 1997), and produced larger eggs (Azevedo et al. 1996). Temperature appears to be a major selective pressure because the same suite of phenotypes arose during experimental evolution in the laboratory (Azevedo et al. 1996; James and Partridge 1995; Neat et al. 1995; Partridge et al. 1994). Phenotypes that evolved at low temperature conferred greater longevity and fecundity in a cold environment, but conferred no noticeable beneﬁts in a warm environment (Bochdanovits and de Jong 2003; McCabe and Partridge 1997; Reeve et al. 2000). Contrary to Stelzer’s idea, egg size did not consistently affect size at maturity, but large eggs did enhance rates of development and growth (Azevedo et al. 1997). Thermal adaptation produced a suite of life-history phenotypes, apparently involving some functional linkage. Despite the evidence that life-history traits coevolve, current theory cannot predict how the entire life history will vary over thermal clines. This shortfall stems from a focus on the evolution of individual traits (or pairs of traits) rather than on the integration of these traits. For example, some models address the effect of size-speciﬁc or stage-speciﬁc mortality on the optimal rate of growth and development, but they do not consider how growth should affect the evolution of reproductive strategies (Cichoñ and Kozłowski 2000; Conover 1998; Wieser 1994). Alternatively, models of the optimal size at maturity explore the consequences of tradeoffs between current and future reproduction, but they do not link strategies of maturation to the allocation of energy among offspring (Berrigan and Charnov 1994; Kozłowski et al. 2004; Sibly and Atkinson 1994). Finally, models of reproductive allocation consider the tradeoffs between offspring quality and quantity, but they do not link reproductive strategies to the evolution of juvenile growth and development (Perrin 1988; Yampolsky and Scheiner 1996). Although much progress has been made by studying speciﬁc tradeoffs and isolated traits, major strides

180

T H E R M A L A D A P TAT I O N

could be made by developing theories that explore the integration of traits throughout ontogeny. To transition from current models that atomize the life history to new models that take a holistic approach, we must embrace the concept of phenotypic integration. As Pigliucci (2003) suggests, think of phenotypic integration as the genetic correlations among traits driven by their functional relationships. Traits such as offspring size and growth rate should be tightly integrated, while those such as offspring size and adult size might be less integrated. For many ectotherms, we could describe the integration of life-history phenotypes as follows: individuals in colder environments grow slower, mature later at a larger size, and produce fewer, larger offspring. This pattern of phenotypic integration describes covariation among individuals resulting from either genetic divergence or thermal plasticity. Although the pattern does not apply to all species,

it occurs frequently enough to warrant our attention. We can interpret this phenotypic integration as a product of constraint or a product of adaptation. If natural selection causes patterns of phenotypic integration, we are faced with the challenge of developing a theory that describes the coadaptation of phenotypes. Such a theory must identify the behavioral and physiological processes that span ontogenetic stages and the tradeoffs caused by these processes. The degree to which adaptation will involve changes in growth and developmental rates, age and size at maturity, or offspring size and number should depend on the relative costs of a change in each trait. But coadaptation involves more than just life-history traits; strategies of thermosensitivity, thermoregulation, and thermal acclimation must also coevolve with the life history. In the next chapter, we shall directly tackle the problem of thermal coadaptation.

CHAPTER 7

Thermal Coadaptation

7.1 Traits interact to determine ﬁtness In the preceding chapters, we covered some major issues in evolutionary thermal biology, namely adaptations of thermal sensitivity, thermoregulatory behavior, thermal acclimation, and life history. As would most evolutionary biologists, we addressed these issues by atomizing the phenotype into mathematically tractable pieces. This approach assumes that natural selection shapes each trait independently. In Chapter 3, we considered the evolution of thermal sensitivity, but assumed organisms do not thermoregulate (or do so independently of their thermal sensitivities). In Chapter 4, we considered the evolution of thermoregulation while assuming organisms have a ﬁxed thermal sensitivity. In Chapter 5, we examined the evolution of thermal acclimation, again assuming no thermoregulatory behavior. Finally, in Chapter 6, we examined the evolution of the life history by assuming ﬁxed thermal sensitivities of survivorship, growth, and development. In each case, these simplifying assumptions enabled us to derive predictions with relative ease. But these predictions should be viewed with caution because real organisms do not evolve under such constraints. Real organisms possess a suite of strategies to cope with thermal heterogeneity, and these strategies undoubtedly evolve according to their combined inﬂuence on ﬁtness. In other words, the behavior, physiology, and life history of a species coadapt to the environment. The concept of coadaptation applies perfectly well to both the genotype and the phenotype. For example, a set of genes will coevolve if their interaction confers greater ﬁtness than other possible sets of genes (reviewed by Templeton 1986; Whitlock et al. 1995). Alternatively, one can focus on the coadaptation of traits controlled by

these genes; two or more traits will coevolve if they enhance ﬁtness through an interaction (Lande 1984; Wolf and Brodie 1998). Two examples of the coadaptation of traits include behavioral drive and behavioral inertia (Huey et al. 2003). In behavioral drive, the plasticity or evolution of behavior drives subsequent evolution of physiology, morphology, or life history. In behavioral inertia, the behavior of an organism ameliorates the stress of thermal heterogeneity, thereby limiting the evolution of other traits. Phenomena such as these suggest that evolutionary biologists are looking at the integrated superstrategies of organisms, whether they care to recognize it or not. Why think in terms of coadaptation when modeling evolution? By constraining some traits and not others, we will miss selective pressures for the correlated evolution of interacting traits. For example, the thermal sensitivity of performance evolves according to the body temperatures experienced by organisms (Chapter 3). An organism can use behavior and physiology to regulate its body temperature, but the beneﬁt of thermoregulation depends on the thermal sensitivity of performance (Chapter 4). Hence, selection should favor some combination of thermal sensitivity and thermoregulatory behavior (Huey and Bennett 1987). An organism that thermoregulates very precisely would gain little advantage from a wide performance breadth. On the other hand, an organism that thermoregulates poorly might need a wide performance breadth to survive and reproduce. Other combinations of phenotypes should also confer particularly high ﬁtness when compared with alternatives. As we shall see, the correlated evolution of these traits can differ qualitatively from the evolution of each trait in isolation. Therefore, a general theory of thermal adaptation must be a theory of thermal coadaptation. In this chapter, I hope to 181

T H E R M A L A D A P TAT I O N

convey some of the more interesting insights that have emerged from this perspective, starting with the coadaptation of pairs of traits and ending with the coadaptation of thermoregulatory behavior, thermal physiology, and life history.

7.2 Coadaptation of thermal sensitivity and thermal acclimation In previous chapters, we studied the evolution of performance curves and their capacity to acclimate to thermal heterogeneity. To better understand the evolution of performance curves, we applied an optimality model developed by Lynch and Gabriel (1987). This model describes the ﬁtness landscape for the thermal optimum and performance breadth, but assumes these parameters cannot acclimate to a changing environment. In Chapter 5, we moved on to consider a model of thermal acclimation developed by Gabriel (1999). This model deﬁnes the optimal shifts in the thermal optimum and performance breadth, given an initial performance curve and the expected thermal change. Given the interaction between the thermal sensitivity and its acclimation capacity, these two traits surely evolve in synchrony. In this section, we shall use a more general model by Gabriel (2005) to determine the optimal combination of thermal sensitivity and thermal acclimation. As we shall see, this model of coadaptation provides unique insights about the evolution of thermal physiology. As in his previous work, Gabriel (2005) modeled a performance curve for survivorship. This curve was determined by a Gaussian function, with the parameters z1 and z2 deﬁning the thermal optimum and performance breadth, respectively (see eqn 3.5). The ﬁtness of the organism depends on the product of performance throughout its life. Thus, an organism dies if its temperature falls outside the range of temperatures encompassed by its performance curve. Given these considerations, what should the performance curve look like in a changing environment? Should the organism possess a wide performance breadth, to tolerate the full range of body temperatures experienced during its life? Or should it possess a narrower performance breadth and shift its thermal optimum as its temperature changes?

The answer depends on both the variability and predictability of the environment. Imagine an environment that switches between two operative temperatures: T1 and T2 . At each transition, the organism knows its body temperature will change, but does not necessarily know the exact magnitude of this change; in other words, the organism has incomplete information. The pattern of thermal change and the phenotypic consequences of acclimation follow those of Gabriel’s original model (Fig. 7.1). The environment switches from T1 to T2 after a period of time (a). Yet, the organism requires some time (b) to acclimate to the thermal change. After some additional time at T2 (c), the environment switches back from T2 to T1 . Following this reversal in temperature, deacclimation requires some time (d). During the time lags (b and d), the performance of the organism suffers. We can interpret this model as a single cycle of thermal change during the lifespan or multiple cycles throughout the lifespan. For either interpretation, we need to know only the total time that an organism spends in each of the four phases: ta , tb , tc , and td . If we assume all genotypes have the same fecundity, ﬁtness (w) scales proportional to the product of No acclimation Acclimation Performance

182

a

b

c

d

a

Time Figure 7.1 Acclimation improves performance in a ﬂuctuating environment. After some period (a), an abrupt shift in environmental temperature cues acclimation, but an organism requires time to respond (b). After some additional period (c ), the environment switches back to its original temperature. However, deacclimation also requires time (d ). The performance of an acclimating genotype suffers during the time lags (b and d ), but the performance of a nonacclimating genotype suffers longer during thermal change (b and c ). Adapted from Gabriel (1999) in The Ecology and Evolution of Inducible Defenses (Harvell, C. D. and R. Tollrian, eds.) by permission from Princeton University Press. © 1999 Princeton University Press.

T H E R M A L C O A D A P TAT I O N

performance during all phases: w ∝ Pta · Ptb · Ptc · Ptd .

(7.1)

Because performance depends on the operative temperature and the thermal sensitivity, we can rewrite eqn 7.1 as w ∝P(T1 , z1ab , z2ab )ta · P(T2 , z1ab , z2ab )tb · P(T2 , z1cd , z2cd )tc · P(T1 , z1cd , z2cd )td

(7.2)

where z1ab and z2ab equal the thermal optimum and performance breadth during ta and tb , and z1cd and z2cd equal the thermal optimum and performance breadth during tc and td . Using eqn 7.2, we can determine the thermal optima and performance breadths that maximize ﬁtness, given time lags and environmental stochasticity. The conclusions drawn here follow from Gabriel’s analysis (Gabriel 2005); additional insights can be found in other analyses of this model (Gabriel 2006; Gabriel et al. 2005). The capacity for acclimation alters the optimal thermal sensitivity throughout the life of the organism. With no capacity for acclimation, selection favors a thermal optimum equal to the expected temperature and a performance breadth proportional to the thermal variance (see Chapter 3). With an unlimited capacity for acclimation, the thermal optimum should instantaneously switch between T1 and T2 as the environment changes; with such capacity, the performance breadth should be as small as physically possible to minimize the impact of a specialist–generalist tradeoff. In reality, time lags for acclimation and deacclimation prevent this ideal response to thermal change. A time lag for acclimation creates a tradeoff between performances during ta and tb . If the thermal optimum equaled T1 during ta , the organism would maximize its performance during ta but suffer poor performance during tb . Alternatively, if the thermal optimum equaled T2 during ta , the organism would suffer poor performance during ta but maximize its performance during tb . A similar tradeoff between performances during tc and td occurs because of the time lag for deacclimation. When time lags exist, the optimal strategy represents a compromise imposed by these tradeoffs. The thermal optimum should always lie somewhere between T1 and T2 , such that the optimal genotype

183

never maximizes its performance in the current environment but always reduces its loss of performance in the future environment. When the operative temperature equals T1 , the thermal optimum that maximizes ﬁtness (z1∗ ) can be calculated as follows: ab tb ∗ . (7.3) z1 = T1 + (T2 − T1 ) ab ta + t b Intuitively, an increase in the time required for acclimation (and thus tb ) should move the thermal optimum during ta closer to the temperature of the future environment, T2 . Likewise, an increase in the time required for deacclimation (and thus td ) should move the thermal optimum during tc closer to T1 : tc . (7.4) z1∗ = T1 + (T2 − T1 ) cd tc + t d Time lags also favor the evolution of wider performance breadths, which enhance performance during acclimation and deacclimation. The optimal performance breadths (z2∗ and z2∗ ) scale ab cd proportional to the magnitude of thermal change (T2 − T1 ). If time lags become sufﬁciently long, an interesting phenomenon emerges. When the time spent acclimating to thermal change exceeds the period of beneﬁt (i.e., tb > tc and td > ta ), selection favors a thermal optimum closer to the future environmental temperature than the current one! Figure 7.2 illustrates this phenomenon for genotypes with incomplete and complete information about the magnitude of change. In the rare event that the periods of acclimation and deacclimation equal the periods of beneﬁt (i.e., ta = tb = tc = td ), selection favors no acclimation of the performance curve. In this special case, the optimal performance curve corresponds to that predicted by a model of evolution without acclimation (e.g., Lynch and Gabriel 1987): the thermal optimum should equal the expected temperature and the performance breadth should scale proportional to the variance of temperature. In Gabriel’s model of coadaptation, only this special case favors a perfect match between the thermal optimum and the expected temperature. All other conditions favor a constant mismatch between the two, which differs qualitatively from the predictions of models that ignore coadaptation (see Chapters 3 and 5). Coadaptation also depends on the predictability of thermal change. We can reasonably assume that

T H E R M A L A D A P TAT I O N

(a)

–1

Acclimation of z1 Short time lag (td) Incomplete information

1

(b)

–1

2

(c)

Fitness

Fitness

184

Acclimation of z1 Long time lag (td) Incomplete information

1

2

(d) Acclimation of z1 and z2 Short time lag (td) Complete information

–1

1

2

Temperature

Acclimation of z1 and z2 Long time lag (td) Complete information

–1

1

2

Temperature

Figure 7.2 The optimal acclimation of performance curves depends on the time lag for physiological change. In each plot, the optimal performance curves at T1 and T2 are depicted by the long-dashed line and the solid line, respectively. For comparison, the short-dashed line depicts the optimal phenotype without acclimation. The thermal axis is relative, with zero being equal to T1 . Optimal strategies are shown for conditions in which individuals have either incomplete or complete information about the thermal change (a and b vs. c and d). In all cases, the duration ofT1 was three times as long as the duration of T2 (ta + td = 0.75). Therefore, when deacclimation requires a very long time (b and d), the optimal strategy of acclimation prepares the organism for the future temperature rather than maximizes performance at the current temperature (i.e., the thermal optimum during T1 lies closer to T2 and the thermal optimum during T2 lies closer to T1 ). Adapted from Gabriel (2005) with permission from Blackwell Publishing.

organisms know of a change in temperature without knowing the exact magnitude of the change. In natural environments, stochasticity of diel and seasonal changes seems more the rule than the exception (see Chapter 2). To incorporate this stochasticity, Gabriel characterized T1 and T2 by their means and variances. Two results emerged from this modiﬁcation of the model. Stochasticity should not affect the thermal optimum at any phase of the organism’s life. For any degree of uncertainty, the thermal optimum shifts according to the long-term mean of thermal change (see also Chapter 5). Nonetheless, unpredictable values of T1 and T2 do affect the optimal performance breadths. As stochasticity increases, the performance breadth should increase. In fact, the selective advantage of thermal acclimation diminishes as the variance of temperature increases (Fig. 7.3). In the rare situation where the environment cycles predictably, the optimal performance breadths before and after acclimation depend only on time lags for shifting the thermal optimum.

The unique predictions offered by Gabriel’s model may help us to better understand the diversity of performance curves observed in nature. Although we currently know little about the time lags for acclimation (see Chapter 5), such constraints could explain why certain species exhibit thermal optima that deviate from their mean body temperatures. For example, Smith and Hadley (1974) repeatedly observed an imperfect match between the environmental temperature and the thermal optimum for photosynthesis in four populations of the Labrador tea (Ledum groenlandicum). After raising plants in a common environment for several years, these investigators exposed the plants to repeated shifts in environmental temperature. Initially, some plants were exposed to a diel cycle of 15◦ C light–10◦ C dark, while others were exposed to 30◦ C light–25◦ C dark. After 12–16 weeks, the environments of these groups were switched; plants in the cool environment were shifted to the warm environment, and vice versa. Two weeks later, both groups were switched back

T H E R M A L C O A D A P TAT I O N

(a)

(b) tb = 0.01

1.4 Relative fitness

185

tb = 0.01 3

tc = 0.2

td = 0.01 cvmin = 0

tdmin = 0

1.2

tdmax

cvmax = 1

2

= 0.3

1 1 0

2 4 Coefficient of thermal variation (cv)

(c)

0.3 0.6 Duration of acclimatory benefit (tc)

(d) td = 0.005

1.4 Relative fitness

0

tc = 0.2

1.4

cv = 0.5

cv = 0.8 1.2

tcmin = 0.1

1.2

tbmin= 0 tbmax= 0.2

tcmax = 0.5 1

1 0

0.1 0.2 Time lag for acclimation (tb)

0

0.2 0.4 0.6 Time lag for deacclimation (td)

Figure 7.3 The selective advantage of reversible acclimation of the thermal optimum depends on properties of the environment and the organism. In each plot, lines depict the selective advantage when a particular parameter was varied in constant steps between minimal (min) and maximal (max ) values. The ﬁtness of an acclimating phenotype was scaled relative to that of a ﬁxed phenotype. Relative ﬁtness depends on the following conditions: (a) the unpredictability of thermal change; (b) the duration of acclimatory beneﬁt (tc ), assuming a brief time lag for acclimation (tb ); (c) the time lag for acclimation (tb ), assuming a brief time lag for deacclimation (td ); and (d) the time lag for deacclimation (td ). Adapted from Gabriel (2005) with permission from Blackwell Publishing.

to their initial environments. Prior to each shift, the investigators measured the thermal sensitivity of the net photosynthetic rate. Based on Gabriel’s model, acclimation and deacclimation should have resulted in thermal optima that lay somewhere between the two diurnal temperatures (15◦ C and 30◦ C). Consistent with this prediction, thermal optima of all populations shifted from 25◦ C in the warm environment to 20◦ C in the cool environment (Fig. 7.4). When exposed to the reverse order of temperatures, only two of the four populations followed this pattern; the other populations exhibited a thermal optimum of 25◦ C in both environments. Possibly, these mismatches between the thermal optimum and the environmental temperature stemmed from an insufﬁcient period for acclimation. But this explanation seems unlikely because 12–16 weeks of acclimation to the same thermal environments did not produce

a closer match between thermal physiology and environmental temperature. Coadaptation would also create geographic variation in performance curves. Environments at high latitudes are colder and vary more than environments at low latitudes. Because of differences in seasonality, temperatures during the winter differ far more along latitudinal clines than do temperatures during the summer (see Fig. 2.8). Simple models of thermal sensitivity predict a decrease in the thermal optimum and an increase in the performance breadth with increasing latitude (see Chapter 3). However, thermal acclimation would enable an organism to specialize for the mean temperature of each season. Latitudinal clines in the seasonality and stochasticity of temperature should affect patterns of thermal physiology. At high latitudes, we should ﬁnd genotypes that shift their

186

T H E R M A L A D A P TAT I O N

5

Net photosynthetic rate (mg CO2 g–1 h–1)

4 3 2

4 3 2 1 10

15

20

25

30

35

15

20

25

Temperature (°C)

thermal optimum between seasonal extremes but retain relatively wide performance breadths in each season. At low latitudes, we should ﬁnd genotypes with relatively constant thermal optima and narrow performance breadths. If we examined performance curves in the winter, we should see a strong latitudinal cline in the thermal optimum for performance. However, the same cline should be much weaker in the summer. One way to evaluate Gabriel’s model of coadaptation would be to see whether latitudinal clines in thermal physiology actually differ between seasons.

7.3 Coadaptation of thermal physiology and thermoregulatory behavior Given the ideas about coadaptation laid out in the preceding section, we can now go one step further and ask how the interaction of thermal physiology and thermoregulatory behavior inﬂuences adaptation. In Chapter 4, we saw that the optimal body temperature depends strongly on the shape of the performance curve. At the same time, the optimal performance curve depends on the body temperatures experienced by an organism. The union of these two results demands the correlated evolution of thermal sensitivity and thermal preference; an evolutionary shift in the thermal optimum should provide selective pressure for a subsequent shift in the preferred temperature, and vice versa (Huey and

30

35

Figure 7.4 Thermal sensitivities of net photosynthetic rate acclimated reversibly to cool environments (open symbols) and warm environments (ﬁlled symbols). Each plot shows data for a population of Ledum groenlandicum. The arrows mark the temperatures of the two acclimation environments during photophase. Adapted from Smith and Hadley (1974) with permission from The Institute of Arctic and Alpine Research.

Bennett 1987). A patchy environment can even cause the evolution of a genetic correlation between physiology and behavior (Nosil et al. 2006). This coadaptation should produce a strong correlation between the thermal optimum and the preferred temperature within and among populations. Evidence of the coadaptation of physiology and behavior comes from two sources: covariations of phenotypes within individuals and among species (reviewed by Angilletta et al. 2002a). For a coadapted genotype, a parallel shift in thermoregulatory behavior should accompany acclimation of a performance curve. For example, when acclimation to a cold environment involves a decrease in the thermal optimum, we should expect a corresponding decrease in the preferred temperature. Only a few teams have investigated the joint acclimation of thermal physiology and thermoregulatory behavior, but their work has generated a range of ﬁndings. Crocodiles maintained at 20◦ C underwent a perfect shift in their thermal optimum for swimming and a small parallel shift in their preferred body temperature (see Fig. 5.3a; Glanville and Seebacher 2006). In contrast, newts held at 25◦ C shifted their thermal optimum for swimming upward and their preferred body temperature downward (Fig. 7.5). Acclimation sometimes involves parallel shifts in the critical thermal limits of performance and the preferred body temperature. Prawns (Macrobrachium acanthurus) maintained at 20◦ C and those maintained at 30◦ C differed in their

T H E R M A L C O A D A P TAT I O N

30

120

15°C

100

20

80

15

60

10

40 CTmax

5

20 0

0

5

10

15

20

25

30

35

40

30

120

25°C

25

100

20

80

15

60

10

40 CTmax

5 0

0

5

10 15 20 25 30 Preferred temperature (°C)

35

40

Relative performance (%)

Number of observations

25

0

187

20 0

Figure 7.5 Acclimation of preferred body temperature (bars) and thermal sensitivities for swimming (solid lines) and running (dashed lines) did not support the hypothesis of coadaptation. After exposure to 15◦ C, the mean preferred temperature and the thermal optima for locomotor performances were approximately 25◦ C. After exposure to 25◦ C, the preferred temperature was much lower than the thermal optima. Vertical dotted lines denote the lower and upper quartiles of preferred body temperature. Adapted from Gvoždík et al. (2007) with permission from Blackwell Publishing.

preferred body temperatures by 7◦ C, but only differed in their critical thermal minima and maxima by 4◦ C (Diaz et al. 2002). The preferred body temperatures and critical thermal limits of grass lizards were positively correlated with acclimation temperature, but acclimation responses were small relative to thermal change; speciﬁcally, the preferred body temperature, critical thermal minimum, and critical thermal maximum of lizards kept at 35◦ C exceeded those of lizards kept at 20◦ C by only 3◦ C (Yang et al. 2008). Acclimation of angelﬁsh to constant temperatures ranging from 20◦ C to 32◦ C affected the critical thermal maximum, but did not affect the preferred body temperature (Pérez et al. 2003). The diversity of responses observed in just these few studies justiﬁes further development of the theory. Interspeciﬁc comparisons have also provided mixed evidence of coadaptation. Dawson (1975) ﬁrst

examined interspeciﬁc correlations between preferred body temperatures and the thermal optima for physiological functions, including heart rate, muscle contraction, and enzyme activity. His review was restricted to reptiles, but established a qualitative match between behavior and physiology in many species (see, for example, Fig. 7.6). Beitinger and Fitzpatrick (1979) concluded the same was true for ﬁsh. Since these seminal reviews, two teams have conducted more rigorous tests of coadaptation at the interspeciﬁc level. Not only did their studies include quantitative analyses of correlated evolution, they also included efforts to control for phylogenetic effects. Both studies were focused on the locomotor performance of lizards; this focus was likely motivated by the observation that many active lizards maintain body temperatures that enable near maximal running speeds (Hertz et al. 1988).

188

T H E R M A L A D A P TAT I O N

(a) 40

(b)

36

Topt for contraction (°C)

Topt for contraction (°C)

40

32

28

24

36

32

28

24

28

30

32 34 36 38 Preferred body temperature (°C)

40

22

24

26 28 30 32 34 Preferred body temperature (°C)

36

38

Figure 7.6 Among species of lizards, the thermal optimum for muscle contraction correlated with the preferred body temperature. Data are isometric twitch tensions (a, Dawson 1975) and tetanic tensions (b, John-Alder and Bennett 1987). Error bars show the optimal range of temperatures for contraction. Dashed lines show where the points should lie if the preferred temperatures perfectly matched the thermal optima.

Huey and Bennett (1987) conducted the ﬁrst phylogenetic comparative analysis of coadaptation. They examined the correlated evolution of the preferred body temperature and the thermal sensitivity of sprinting in 12 species of Australian skinks. Because they did not know the timing of divergence among taxa, they chose a minimum-evolution approach in which ancestral phenotypes were estimated by averaging extant phenotypes. Then, they regressed hypothetical evolutionary changes in the thermal optimum, critical thermal minimum, and critical thermal maximum against changes in the preferred body temperature. This analysis supported the conclusion that the thermal optimum and the critical thermal maximum evolved in synchrony with the preferred body temperature. An analysis of raw variables also supported this conclusion. But this story has an interesting twist. Just a few years later, the emergence of a new phylogenetic hypothesis prompted Huey and Bennett to reanalyze their data. With the help of Garland, they used the new phylogeny to conduct a variety of phylogenetic analyses, including correlations of independent contrasts (Garland et al. 1991). The new analyses suggested only partial coadaptation had occurred. The preferred body temperature appears to have evolved in synchrony with the critical thermal maximum for sprinting, but did not evolve in synchrony with the critical thermal minimum or the thermal optimum (Fig. 7.7).

A similar study by Bauwens and colleagues (1995) provided stronger evidence of coadaptation. These investigators measured the preferred body temperatures and the thermal sensitivities of sprinting in 13 species of lacertid lizards. Analyses of independent contrasts showed that preferred body temperature covaried positively with the critical thermal maximum and the thermal optimum (Fig. 7.8). Moreover, the precision of thermoregulation and the breadth of performance coevolved very tightly, but in a way that contradicts logic: species with narrower performance breadths thermoregulated less precisely! As in Australian skinks, thermoregulatory behavior and thermal physiology seems only partially coadapted in lacertids. These phylogenetic analyses tell us whether behavior and physiology have evolved in the same direction, but they do not tell whether traits have coadapted perfectly. This problem stems from an inherent limitation of the phylogenetic method. Independent contrasts deﬁne the deviation in the value of a trait between two taxa since they diverged from a common ancestor (see Box 1.5). For instance, we would calculate an independent contrast for the thermal optimum by subtracting the thermal optimum for one species from that of its sister species. Unfortunately, this contrast retains no information about the absolute value of the thermal optimum in each species. The same would be true of independent contrasts for preferred body temperature.

T H E R M A L C O A D A P TAT I O N

Thermal 0.21/0.21 Relative optimum sprint speed for sprinting

36

r = 0.58

26 24

2* 0 .6 0*/

Critical thermal maximum

0.64*/0.64*

28

0.45/0.51*

65*

30

0 .6

32

0.61 */0.

Thermal physiology

34

Thermal performance breadth

0.5

8/0

.58

–0.90**/–0.90**

(b)

3

Standardized contrast for Topt

22

2

22

24 26 28 30 32 34 Preferred body temperature (°C)

Thermoregulatory behavior

Thermal optimum for sprinting (°C)

(a)

189

36

r = 0.34

1

Breadth of 2 preferred –0.3 .31/ temperatures –0 0.48/0.50 Median preferred temperature

Figure 7.8 Relationships among aspects of thermal physiology and thermoregulatory behavior in lacertid lizards support the hypothesis of coadaptation. Correlation coefﬁcients (double-headed arrows) or standardized regression coefﬁcients (single-headed arrows) were calculated using independent contrasts; two values are reported for each relationship because two plausible phylogenetic hypotheses were available. Asterisks denote statistically signiﬁcant coefﬁcients. Adapted from Bauwens et al. (1995) with permission from Blackwell Publishing.

0

–1

–2 0.0

Relative sprint speed

0.5

1.0 1.5 2.0 2.5 Standardized contrast for Tp

3.0

Figure 7.7 The thermal optimum for sprinting did not coevolve with the preferred body temperature of Australian skinks. The signiﬁcant correlation between raw variables (a) indicates coadaptation, but the absence of a correlation between independent contrasts (b) suggests the relationship between raw variables resulted from phylogenetic pseudoreplication. Data are from Garland et al. (1991).

Therefore, a correlation between contrasts cannot tell us whether the preferred body temperature matches the thermal optimum for performance. To determine the degree of coadaptation, we must examine the correlation between raw variables. In doing so, we discover something interesting: the thermal optimum for sprinting usually exceeds the preferred body temperature (Fig. 7.7a; Bauwens et al. 1995; Huey et al. 1989). This observation has

two possible interpretations. Either some unidentiﬁed mechanism constrains coadaptation, or perfect coadaptation does not necessarily imply the preferred temperature should equal the thermal optimum. As we shall see, recent modeling points toward the latter interpretation.

7.3.1 Mechanisms favoring a mismatch between preferred temperatures and thermal optima Despite the current dogma, coadaptation should not always produce a perfect match between the preferred temperature and the thermal optimum. Two distinct conditions can favor a mismatch. First, when individuals cannot thermoregulate with perfect accuracy, the thermal optimum should exceed

T H E R M A L A D A P TAT I O N

w=

CT max

f (Tb ) · P(Tb ),

(7.5)

Tb =CTmin

where f (Tb ) equals the frequency distribution of body temperatures. In biological terms, ﬁtness equals the sum of performance over time, such as the case when performance directly affects fecundity (see Box 3.3). Both symmetrical and asymmetrical functions were examined for the frequency distribution, f (Tb ), and the performance curve, P(Tb ). For an asymmetrical performance curve, the thermal optimum exceeded the mean body temperature that maximized ﬁtness (Fig. 7.10). The more an organism’s temperature varied, the greater the optimal mismatch between the preferred temperature and the thermal optimum.13 This effect was particularly 13

The same phenomenon was predicted by earlier models of thermal adaptation (Beuchat and Ellner 1987; Gilchrist 1995).

(a)

Performance

the preferred temperature. Second, when warmadapted organisms outperform cold-adapted organisms, the thermal optimum should also exceed the preferred temperature. Let us brieﬂy examine the reasons for these conclusions. Martin and Huey (2008) developed a simple model to show how the accuracy of thermoregulation affects the course of coadaptation. If an organism thermoregulates with perfect accuracy, the greatest performance obviously results from maintaining a body temperature that equals the thermal optimum. But what mean body temperature maximizes performance when an organism thermoregulates imperfectly? The answer to this question depends not only on the variation in body temperature but also on the shape of the performance curve. For a symmetrical performance curve, unusually high body temperatures and unusually low body temperatures have equal impacts on performance (Fig. 7.9a). For an asymmetrical performance curve, unusually high body temperatures impair performance more than unusually low temperatures do (Fig. 7.9b). Because performance curves tend to be asymmetrical, inaccurate thermoregulation should affect the optimal preferred temperature. To illustrate this point, Martin and Huey found the mean body temperature that maximized the following ﬁtness function:

(b)

Performance

190

Temperature Figure 7.9 The effect of thermal variation on performance depends on the shape of the performance curve. For symmetric curves (a), body temperatures greater than and less than the thermal optimum lead to equal performances. For asymmetric curves, a body temperature below the thermal optimum enables better performance than does a body temperature above the thermal optimum. Based on Martin and Huey (2008).

exaggerated for thermal specialists because deviations from the thermal optimum imposed severe reductions in ﬁtness (Fig. 7.10). However, both specialists and generalists should prefer temperatures below their thermal optima given extreme asymmetry of the performance curve. To evaluate the predictions of their model, Martin and Huey conducted a phylogenetic comparative analysis of coadaptation. For 52 species of lizards, they calculated the difference between the mean preferred temperature and the thermal optimum for sprinting. The asymmetry (A) of each performance curve was characterized by In Fig. 3.18, compare the optimal performance curves predicted by the levels of thermal heterogeneity. As thermal variation increases among generations, selection favors a greater difference between the mean environmental temperature and the thermal optimum.

(b) Relative fitness

1.0 Relative fitness

(a)

0.8 0.6 0.4 0.2 0

1.0

(c)

0.8

Optimal mismatch (°C)

T H E R M A L C O A D A P TAT I O N

0.6 0.4 0.2 0

0 20 40 Body temperature (°C)

191

3 2 1 0

0 20 40 Body temperature (°C)

0

0.4 0.8 Asymmetry

Figure 7.10 The optimal mismatch between the preferred temperature and the thermal optimum depends on the asymmetry and breadth of the performance curve. (a and b) Examples of symmetries and breadths used to model performance curves. (c) The three lines correspond to the breadths shown in (b). The optimal mismatch increases with increasing asymmetry but decreases with increasing breadth. Adapted from Martin and Huey (2008, The American Naturalist, University of Chicago Press). © 2008 by The University of Chicago.

(b)

8 6 4 2 0 –2

Topt – Tp

Topt – Tp

(a)

8 6 4 2 0 –2 6 8 10 12 14 16 Performance breadth

(d)

4 2 0 –2 –4

Contrast for Topt – Tp

(c)

Contrast for Topt – Tp

0.2 0.4 0.6 Asymmetry 4 2 0 –2 –4

–0.15 0.00 0.15 Contrast for asymmetry

–2 0 2 4 6 Contrast for breadth

the following function: A=

2Topt − CTmax − CTmin CTmax − CTmin

(7.6)

The empirical data were qualitatively consistent with the model (Fig. 7.11). First, the observed mismatch between the preferred temperature and the thermal optimum increased as the asymmetry of the performance curve increased. Second, thermal specialists exhibited a greater mismatch than did thermal generalists. Analyses of raw data and independent contrasts showed similar patterns. However, close inspection of the raw data reveals that the observed mismatches greatly exceed those predicted by the model (compare Figs 7.10c and 7.11a).

Figure 7.11 The observed mismatch between the preferred body temperature and the thermal optimum for sprinting depends on the asymmetry and breadth of the performance curve. As predicted, the mismatch increases as asymmetry increases and decreases as breadth increases. (a and b) Raw data. (c and d) Independent contrasts. Adapted from Martin and Huey (2008, The American Naturalist, University of Chicago Press). © 2008 by The University of Chicago.

Therefore, some of this variation likely reﬂects the evolutionary consequences of other processes. A possible explanation relates to the ﬁtness function used in the model. By calculating ﬁtness as the sum of performance, Martin and Huey assume that the body temperature of the organism never exceeds its thermal limit for survival. Relaxing this assumption would favor a greater mismatch between the thermal optimum and the preferred temperature (Martin and Huey 2008). But some of this mismatch might stem from a mechanism that has nothing to do with the asymmetry of performance curves. A second mechanism that promotes a mismatch between the preferred temperature and the thermal optimum involves thermodynamic constraints on

192

T H E R M A L A D A P TAT I O N

(a)

Performance

Chapter 3), except we assumed the mean performance of the organism depended on its thermal optimum: w=

CT max

fg (Tb ) · P(Tb , Topt ),

(7.7)

Tb =CTmin

where fg (Tb ) equals the frequency distribution of body temperatures in the current generation (compare with eqn 3.13). The thermodynamic effect on performance was modeled according to metabolic theory (Gillooly et al. 2001, 2002; Savage et al. 2004):

(b)

Performance

P ∝ e−E/kTopt ,

Temperature Figure 7.12 (a) The hypothesis that “hotter is better” describes a thermodynamic effect on the performance of an organism. The mean or maximal performance over a range of temperatures increases as the thermal optimum increases. (b) When hotter is better, selection favors a thermal optimum for performance that exceeds the mean body temperature (dashed line). This shift enhances performance at high body temperatures without impairing performance at the mean body temperature.

performance. From Chapter 4, recall that the performance of warm-adapted organisms usually exceeds that of cold-adapted organisms. In other words, hotter is better (Fig. 7.12a). Despite this common observation, models of thermal adaptation assume that adaptation to cold environments yields the same mean performance as adaptation to warm environments. Which predictions, if any, would change if we included a thermodynamic effect on performance? To answer this question, Asbury and I modeled the optimal performance curve under the assumption that hotter is better (D. Asbury and M. J. Angilletta, unpublished observations). We used a model very similar to the one developed by Gilchrist (1995; see

(7.8)

where E equals the mean activation energy of rate-limiting reactions, and k equals Boltzmann’s constant. Because a higher thermal optimum enhances performance, an organism achieves maximal ﬁtness when its thermal optimum exceeds its mean body temperature (Fig. 7.12b). An increase in the variance of body temperature intensiﬁes this effect (Fig. 7.13). Thermodynamic constraints and inaccurate thermoregulation independently promote a mismatch between physiology and behavior, as long as the thermodynamic effect does not change the asymmetry of the performance curve (Martin and Huey 2008). These models assume the thermal sensitivity of a single performance determines the ﬁtness of an organism, but many physiological processes occur simultaneously. These processes likely differ in their thermal sensitivity and their impact on ﬁtness. Variations in thermal optima and performance breadths among processes should inﬂuence coadaptation. In some species, very different performances proceed maximally at a single body temperature (Du et al. 2000; Huey 1982; Ji et al. 1995; Luo et al. 2006; Stevenson et al. 1985). In other species, thermal optima differ among physiological performances (Chen et al. 2003; Ji et al. 1996; Van Damme et al. 1991). An organism can solve this problem by varying its body temperature such that all processes proceed at a maximal rate for some duration. Alternatively, an organism can maintain a body temperature that permits all processes to proceed at a moderate rate. When adopting the latter strategy, the optimal body temperature depends on the relative impact of each process on ﬁtness (see Chapter 4).

T H E R M A L C O A D A P TAT I O N

(a)

focused on one or two physiological performances, whose links to ﬁtness remain unclear.

100 High

80

7.3.2 Predicting coadapted phenotypes

60 Relative performance (% of maximum)

193

Medium

40 20

Low

0 0

10

20

30

40

50

(b) 100 High 80 60 40 Medium

20

Low

0 0

10

20 30 Temperature (°C)

40

50

Figure 7.13 When hotter is better, optimal performance curves deviate from those described in Chapter 3 (D. Asbury and M. J. Angilletta, unpublished observations). Here, optimal performance curves without a thermodynamic constraint (a) are compared with those when hotter is better (b). For each curve, variation in body temperature among generations was high, while variation within generations was low, medium, or high. Variations within and among generations correspond to those shown in Fig. 3.17. As a reference, the dashed line in each plot approximates the mean environmental temperature. When body temperature varies within generations, a thermodynamic effect on performance favors a thermal optimum that exceeds the mean body temperature.

If performance breadths vary among physiological processes, thermoregulatory behavior should coevolve with the most thermally sensitive performances. For example, the preferred temperature of Sceloporus undulatus appears more closely related to the thermal sensitivity of energy assimilation than the thermal sensitivities of sprinting and endurance (Angilletta et al. 2002b). When performances differ in their thermal sensitivity, detecting coadaptation will require extensive knowledge of the relationships between performance and ﬁtness. Keep in mind that empirical studies of coadaptation have

The current theory falls short of providing a full understanding of coadaptation. First, let us review what the theory does address. The theory clearly predicts covariation between thermal physiology and thermoregulatory behavior. We expect the preferred temperature to correlate with the thermal optimum for performance, although variations in the precision of thermoregulation and the sensitivity to high temperatures would weaken this correlation. We also expect the preferred range of temperatures to correlate with the thermal breadth of performance, particularly when performance determines survivorship. These predictions come from the informal integration of models that focus on either thermal physiology or thermoregulatory behavior. Nevertheless, this informal integration cannot predict which combination of strategies should occur in a particular environment. Why do some organisms thermoregulate precisely and tolerate a very narrow range of temperatures? Why do other organisms tolerate a broad range of temperatures and let their body temperature ﬂuctuate with the environmental temperature? To answer these questions, we need a very different theory of coadaptation. This theory must evaluate the costs and beneﬁts of all combinations of thermal sensitivity and thermoregulatory behavior in a range of environments. In the absence of a formal theory, how do we know which combination of strategies confers the greatest ﬁtness? At one extreme, an organism can regulate its body temperature precisely and specialize for physiological function at its target temperature. At the other extreme, the organism can let its temperature ﬂuctuate and evolve a performance curve and an acclimation capacity that suit the variation in body temperature. Somewhere in the middle lie countless combinations of thermoregulatory behavior and thermal physiology. The course of adaptation depends on the costs and beneﬁts of these strategies in the current environment. If the net beneﬁt of a pure behavioral strategy exceeds that of a pure physiological strategy, then some

194

T H E R M A L A D A P TAT I O N

degree of thermoregulation should evolve (and vice versa). Given the many factors that determine the costs and beneﬁts of these strategies (see Chapters 3–5), coadaptation could be a very complex process to understand. Yet, I hold some hope that we can simplify the problem. Coadaptation would be a much simpler process if one strategy conferred an inherently greater beneﬁt or lesser cost than did other strategies. Recently, my colleagues and I drew some conclusions about coadaptation from the belief that the beneﬁt of thermoregulation inherently exceeds the beneﬁts of thermal generalization and thermal acclimation (Angilletta et al. 2006a). This belief rests on two assumptions: (1) a jack of all temperatures is a master of none (see Chapter 3) and (2) hotter is better (see Chapter 4). If a jack of all temperatures really is a master of none, a thermal specialist with a constant body temperature would have a higher ﬁtness than a thermal generalist with any distribution of body temperatures (Gilchrist 1995; Lynch and Gabriel 1987). If hotter is truly better, a specialist adapted to a high body temperature would enjoy greater ﬁtness than a specialist adapted to a low body temperature (Frazier et al. 2006). Combining

Spatial and temporal covariations between temperature and predation risk

Spatial and temporal variations in operative environmental temperature

Spatial and temporal covariations between temperature and other resources

these assumptions, we proposed that the cost of thermoregulation drives coadaptation (Fig. 7.14). Environments that pose little cost of thermoregulation favor specialists that thermoregulate precisely. The optimal body temperature depends not only on the cost of thermoregulation but also on the exact magnitude of the thermodynamic effect and physical constraints on molecular stability. Alternatively, more costly environments favor genotypes that thermoregulate imprecisely or even not at all. The optimal thermal sensitivity of these genotypes depends on the thermal heterogeneity within and among generations and the mathematical relationship between performance and ﬁtness. If ﬁtness depends on the sum of performance over time, a specialist should evolve when body temperature varies relatively little among generations (Gilchrist 1995). A generalist should evolve if body temperature varies greatly among generations (Gilchrist 1995) or if ﬁtness depends on the product of performance over time (Lynch and Gabriel 1987). Importantly, selection might favor generalization through developmental or reversible acclimation (Gabriel 2005; Gabriel and Lynch 1992), which would appear as specialization on a short temporal scale. Predictable variation in body temperature within generations

Reversible acclimation

Specialist when fitness depends on the sum of performance

Thermoregulation

Temporal variation in body temperature

Thermal sensitivity of performance Generalist when fitness depends on the product of performance

Relatively high variation in body temperature among generations

Developmental acclimation

Figure 7.14 The evolution of thermoregulatory behavior should drive the thermal coadaptation of traits. Both abiotic and biotic aspects of the environment determine the optimal degree of thermoregulation, which in turn deﬁnes the variation in body temperature. Variations in body temperature within and among generations deﬁne the selective pressures on thermal sensitivity and its acclimation. This view implicitly assumes that (1) specialists outperform generalists and (2) warm-adapted genotypes outperform cold-adapted genotypes. Adapted from Angilletta et al. (2006a, Physiological and Biochemical Zoology, University of Chicago Press). © 2008 by The University of Chicago.

T H E R M A L C O A D A P TAT I O N

This simpliﬁed view of coadaptation relies on the validity of our two critical assumptions. Current evidence supports the existence of a thermodynamic effect on performance. Indeed, warm-adapted genotypes usually outperform cold-adapted genotypes (i.e., hotter is better; see Chapter 4). Less convincing evidence exists for a universal tradeoff between specialization and generalization; a jack of all temperatures can be a master of all (see Chapter 3). Yet, our current understanding of biochemical adaptation indicates this tradeoff must occur at the biochemical level, even if it can be masked at the organismal level (see Angilletta et al. 2003). If these assumptions ultimately prove incorrect, the process of coadaptation will differ from the one depicted in Fig. 7.14. The absence of thermodynamic constraints or specialist– generalist tradeoffs would weaken selection for thermoregulation and strengthen selection for generalization (see, for example, Gilchrist 1995). Ultimately, biologists may have no choice but to develop more complex models of coadaptation.

7.4 Coadaptation of thermoregulatory behavior, thermal physiology, and life history To achieve either precise thermoregulation or thermal generalization, an organism must invest resources which then become unavailable for other functions. Regulating body temperature in a changing environment requires time or energy and exposes organisms to risks they might otherwise avoid. Tolerating extreme temperatures requires energy to maintain a suite of proteins or to turn over existing proteins during thermal change. Decisions to make these investments undoubtedly shape the survivorship and fecundity of an organism in multiple ways. Therefore, we should expect the thermoregulatory behavior, thermal physiology, and life history of an organism to interact during evolution. Biologists use the concept of a tradeoff to study the interactions between physiological performances and life-history traits (Dunham et al. 1989; Stearns 1992). An acquisition tradeoff occurs whenever an organism incurs some risk of mortality associated with acquiring resources (e.g., energy, mates, or even microclimates). An allocation tradeoff

195

occurs whenever an organism divides a ﬁnite supply of resources among competing demands (e.g., growth vs. reproduction). The inﬂux and division of resources determine life-history traits, such as age-speciﬁc body size, survivorship, and fecundity. Most models of life-history evolution consider relatively simple problems, such as the division of energy between growth and reproduction (see Chapter 6). But organisms must divert incoming energy to many other functions, including maintenance, locomotion, and development. The thermal sensitivity of these performances will depend on the same factors thought to inﬂuence life-history traits. We need consider only the most basic conclusions of life-history theory to forge a connection with thermal adaptation. Classic models have shown the optimal life history depends strongly on agespeciﬁc rates of mortality (Gadgil and Bossert 1970; Hirshﬁeld and Tinkle 1975; Law 1979; Michod 1979). These models distinguish between two types of mortality risk: intrinsic and extrinsic. The intrinsic risk of mortality depends on the organism’s phenotype, whereas the extrinsic risk does not. For example, an animal that thermoregulates behaviorally may draw the attention of predators and hence have a greater intrinsic risk of mortality. Yet, even an inactive animal can fall prey to a predator, which suggests some extrinsic risk also exists. This background risk of mortality should drive the evolution of behavior, physiology, and life history. When the environment poses a high extrinsic risk of mortality, an individual faces low odds of surviving regardless of its phenotype. Hence, natural selection favors individuals that reproduce earlier and more often, even if doing so involves risky behaviors to acquire the necessary resources (Stearns 2000). Let us think about this result in the context of thermal adaptation. As with any performance, the rate of reproduction depends on body temperature. More precise thermoregulation should enhance reproductive performance. Thus, higher extrinsic mortality should cause individuals to accept more risk when seeking preferred microclimates. Extrinsic mortality should also affect the evolution of performance curves. Direct effects stem from the allocation of resources between competing functions (see Fig. 3.30a). For example, mean growth rate over a range of temperatures depends

196

T H E R M A L A D A P TAT I O N

on the allocation of energy between growth and reproduction. If extrinsic mortality increases, the optimal allocation of energy shifts from growth toward reproduction. Consequently, natural selection will decrease the mean growth rate and increase the mean reproductive rate. Size-speciﬁc mortality can have the opposite effect. When an individual must reach a certain size to survive a stressful period, energy should be allocated to growth at the expense of other performances. In this case, natural selection will increase the mean growth rate, resulting in patterns such as countergradient variation (Conover 1992). Indirect effects stem from the coadaptation of thermoregulatory behavior and thermal sensitivity. Because thermoregulation determines the mean and variance of body temperature, high extrinsic mortality not only favors precise thermoregulation but also favors a corresponding change in the thermal optimum and performance breadth (see Section 7.3). Overall, we should expect the evolution of precise thermoregulation, thermal specialization, early maturation, and greater reproduction when the environment poses a high extrinsic risk of mortality.

7.5 Constraints on coadaptation Optimality models deﬁne the ﬁtness landscape, but the dynamics of coadaptation depend on genetic constraints as well as selective pressures (see Chapter 1). Because traits interact to determine ﬁtness, a genetic constraint for one trait can alter the natural selection of other traits (Warner 1980). Some biologists have suggested the primary response to thermal change involves thermoregulatory behavior (Hertz 1981). This behavioral response would then impose selective pressure on thermal physiology (Huey and Bennett 1987), a process referred to as behavioral drive (Huey et al. 2003). Yet little evidence exists to conﬁrm the belief that behavioral traits evolve more readily than physiological or life-historical traits (Artacho et al. 2005; Carlson and Seamons 2008). In fact, far more information exists about the heritability of thermal physiology and life history than exists about the heritability of thermoregulatory behavior. That said, we really do not know all that much about the heritability of thermal physiology (see Chapter 3). Therefore,

we must await further research before drawing general conclusions about the limits to coadaptation imposed by genetic variance. Genetic correlations between traits can either prohibit or facilitate coadaptation (Wolf and Brodie 1998). For example, if genotypes that prefer high temperatures tend to perform best at low temperatures, selection to increase the preferred temperature will decrease the thermal optimum. This negative genetic correlation would prevent coadaptation. Alternatively, a positive genetic correlation would facilitate coadaptation (Box 7.1). Such genetic correlations result from pleiotropic genes, physical linkage, or linkage disequilibrium. Therefore, mutation, recombination, and selection can eliminate existing genetic correlations and generate new ones. Indeed, coadaptation itself can result in the evolution of genetic correlations that facilitate future coadaptation. For example, imagine a metapopulation in which local environments cause a different suite of phenotypes to evolve in each population. Migration among populations will introduce genotypes whose suite of phenotypes differs from the local suite. Even if no physical linkage exists between coadapted alleles, this combination of adaptation and migration would cause a genetic correlation by linkage disequilibrium. The strength of the genetic correlation increases in direct proportion to the rate of migration among populations (Nosil et al. 2006). In this way, selected individuals tend to pass on alleles that promote coadaptation.

7.6 Conclusions In this chapter, a fundamental tenet of the theory of thermal adaptation has been called into question. Simple models either predict or assume a close match between an organism’s mean body temperature and its thermal optimum. Two evolutionary processes can lead to mismatches between the body temperature and the thermal optimum. First, time lags for acclimation and deacclimation favor a mismatch to reduce the loss of performance during thermal change. Second, imprecise thermoregulation favors a mismatch to prevent the loss of performance at high temperatures. These conclusions could have been inferred from models that ignore coadaptation (Gabriel 1999; Gilchrist 1995), but an

T H E R M A L C O A D A P TAT I O N

197

Box 7.1 Evolution of preferred body temperature by indirect selection Experimental evolution of Drosophila melanogaster provided indirect evidence of a genetic correlation between thermoregulatory behavior and thermal physiology. Good (1993) maintained 10 populations of ﬂies at each of three temperatures (25◦ C, 27◦ C, and 30◦ C) for 15 generations. After 10 generations of selection, ﬂies from all lines were raised at 25◦ C and their preferred body temperatures were measured in a thermal gradient. Despite the fact that all generations experienced a uniform environment, preferred body temperatures had diverged genetically among selection lines. Speciﬁcally, ﬂies that evolved at a higher temperature preferred a higher temperature (Fig. 7.15). Because no direct selection for thermoregulation could have occurred without thermal heterogeneity, only two processes

can explain the evolution of preferred body temperature: genetic drift caused by a ﬁnite population or a correlated response to selection of thermal physiology. The latter explanation seems plausible considering that other selection experiments with D. melanogaster documented the evolution of thermal physiology (see Chapter 3). Unfortunately, Good did not determine whether the thermal optimum for any performance had diverged among selection lines. A positive genetic correlation between the preferred temperature and the thermal optimum would facilitate coadaptation, except in cases where selection favors a greater mismatch between these traits (see Section 7.3.1).

70

Percentage of flies

60

25°C 27°C 30°C

50 40 30 20 10 0 Cold (15–21°C) Intermediate (21–27°C) Region of thermal gradient

Hot (27–40°C)

Figure 7.15 The distributions of ﬂies (Drosophila melanogaster) in a thermal gradient differed among lines evolving at constant temperatures (25◦ C, 27◦ C, and 30◦ C). Flies that evolved at 25◦ C preferred the cool region of the gradient, while ﬂies that evolved at 30◦ C preferred the intermediate region. Flies from all lines avoided the hot region. Adapted from Good (1993) with permission from Elsevier. © 1993 Elsevier Ltd.

explicit focus on the interactions among traits makes certain processes more transparent. A formal theory of coadaptation would enable signiﬁcant advances in evolutionary thermal biology. Models that ignore coadaptation predict thermoregulatory behavior, thermal physiology, or life history, but cannot predict the optimal combination of phenotypes. Two constraints on thermal adaptation will facilitate our transition to a general theory: (1) specialist–generalist tradeoffs and (2)

thermodynamic effects. Given these constraints, coadaptation should proceed according to the costs of thermoregulation. Efforts should continue to establish the nature of these constraints in a wide range of organisms. At the same time, we should also develop mathematical models of coadaptation that will enable us to relax these constraints. Finally, we need to exploit a larger set of empirical tools for testing hypotheses about coadaptation. Multivariate selection gradients can reveal interactions between

198

T H E R M A L A D A P TAT I O N

traits that inﬂuence ﬁtness (Brodie et al. 1995), such as the interaction between the preferred temperature and the thermal optimum. Physical or genetic manipulations of thermoregulatory behavior and thermal physiology (e.g., Krochmal and Bakken 2003; Sack et al. 2005) can also be used to assess the

ﬁtness consequences of decoupling coadapted traits. Experimental evolution in the laboratory can reveal the correlated evolution of behavior, physiology, and life history. If we fully embrace a coevolutionary perspective, the theory of thermal adaptation may change radically in the coming years.

CHAPTER 8

Thermal Games

8.1 Filling the ecological vacuum In previous chapters, we implicitly assumed that the thermal adaptation of a species does not depend on the density or frequency of other species. In other words, we have dreamed of evolution in an ecological vacuum. This incredibly naïve approach is slowly being supplanted by a more realistic one: evolutionary games. The theory of evolutionary games deals with the evolution of strategies whose ﬁtness depends on the strategies of other organisms. The concept of an evolutionary game was borrowed from economics and popularized in biology by John Maynard Smith (Maynard Smith 1972, 1976, 1982; Maynard Smith and Price 1973). Unfortunately, the seeds sown by Maynard Smith have grown slowly within certain biological disciplines. Indeed, thermal biology seems untouched by game theory. In this chapter, I seek to stimulate progress by promoting a theory of thermal games. By thermal games, I do not mean the Olympics of temperature; rather, I mean thermal adaptation in an environment where the optimal strategy depends on the strategies of competitors, predators, prey, parasites, and mutualists. Biologists refer to this process as frequency-dependent selection (Nowak and Sigmund 2004). Although game theory usually calls to mind analytical models that deﬁne evolutionarily stable strategies (Maynard Smith 1976), no single approach deﬁnes all of game theory. The current theory includes any model where the ﬁtness of a strategy depends on its frequency and the frequencies of other strategies. Therefore, optimality, quantitative genetic, and allelic models constitute the modern theory of evolutionary games (for a history of this theory see Sigmund 2005).

Why must we consider models of thermal games? Numerous case studies have shown that organisms adjust their phenotype in response to the presence of other organisms. Recall the thermoregulatory behaviors of individuals in the presence and absence of others. Bluegill sunﬁsh occupy their preferred thermal habitat in isolation, but shift to a less preferable habitat when exposed to a larger conspeciﬁc (Medvick et al. 1981). Similarly, we know zooplankton avoid warm shallow waters when they detect the presence of a predatory ﬁsh (Loose and Dawidowicz 1994). Finally, locusts bask more frequently when infected by a fungus than they do in the absence of this pathogen (Ouedraogo et al. 2004). Results such as these expose the need for a theory of thermal adaptation that accounts for the interactions among individuals. But how should we incorporate these interactions into our theory? One approach would be to assume certain behaviors impose costs that arise from the interaction with another species; for instance, we might assume that predators impose an elevated risk of mortality in certain habitats but not others. The current theory can easily be extended to include such assumptions and some models explicitly consider interactions among species (in Chapter 3, see the model by Polo et al. 2005). Still, these models treat other species as inert entities, which cannot alter their own phenotypes (Lima 2002). Game theory avoids this unrealistic assumption, letting us consider the coevolution of species that results from any type of ecological interaction. In this chapter, I focus more on introducing a theory of thermal games than on evaluating its performance. Rigorous tests must wait until more thermal biologists adopt this theory and design experiments to test its predictions.

199

200

T H E R M A L A D A P TAT I O N

8.2 Approaches to the study of frequency-dependent selection The classic approach to understanding evolutionary games rests on the concept of an evolutionarily stable strategy. First, we would translate our assumptions about the game into a set of equations that link the phenotypes of two or more organisms to their respective ﬁtnesses. Next, we would use these equations to ﬁnd a strategy that can invade a population when rare, but cannot be invaded by another strategy once it becomes common (Maynard Smith 1982). In other words, we ﬁnd an evolutionarily stable strategy. Such a strategy would ultimately evolve if a population was adapting to a stable set of abiotic and biotic interactions and possessed sufﬁcient additive genetic variance in the phenotype of interest (see Chapter 1). An evolutionarily stable strategy represents a special case of a more general kind of equilibrium, called a Nash equilibrium. A Nash equilibrium equals a set of strategies for which no other strategy would enhance an individual’s ﬁtness, assuming other individuals maintain their strategies (Riechert and Hammerstein 1983). However, a population of individuals exhibiting a Nash equilibrium can be inﬁltrated by individuals exhibiting an alternative strategy that confers equal ﬁtness; thus, the possibility of genetic drift requires a distinction between a Nash equilibrium and an evolutionarily stable strategy. For a Nash equilibrium to qualify as an evolutionarily stable strategy, either no alternative strategy can confer as much ﬁtness as the Nash equilibrium, or alternative strategies become less ﬁt than the Nash equilibrium as they increase in frequency. Some models predict no evolutionarily stable strategy and others predict several. In practice, an evolutionarily stable strategy could be pure or mixed. Pure strategies exist when all individuals of a population should exhibit the same phenotype. Mixed strategies exist when either (1) proportions of a population adopt different strategies or (2) an individual possesses a set of strategies adopted with certain probabilities (Dawkins 1980). Usually, the evolutionarily stable strategy differs from the optimal phenotype in the absence of frequency-dependent selection. Complementary approaches to the search for an evolutionarily stable strategy involve the use

of analytical models and numerical simulations, including individual-based models (DeAngelis and Mooij 2005). For an analytical model, one must ﬁrst identify whether a rare genotype could invade a large population of other genotypes. Successful invasion suggests the mutant can become established in the population, whereupon it would coexist with or exclude other genotypes. To determine a genotype’s potential for invasion, one calculates the limit of the growth rate of a genotype as its density approaches zero; this limit obviously depends on the resident genotypes and other environmental factors. After establishing a genotype’s ability to invade a population when rare, one must also establish the genotype’s ability to resist invasion when abundant. Thus, an evolutionarily stable strategy prevents the limit of all other genotypes from exceeding the critical value for population growth (Proulx and Day 2001). For more complex models, one cannot solve analytically for an evolutionarily stable strategy. Instead, one must use an individual-based model to simulate the invasion of a rare mutant in a population dominated by another genotype. In such simulations, phenotypes can evolve under constraints imposed by quantitative or multilocus genetics (see Chapter 1). The details of the genetic system and the assumptions about the selection gradients determine the dynamics of coevolution. Individual-based models of evolutionary games can yield stable equilibria that are analogous to the optimal phenotype in a model of frequency-independent selection. In this chapter, we shall consider analytical and individualbased models of thermal games.

8.3 Optimal thermoregulation in an evolutionary game Considering the examples provided in Section 8.1, we should gain considerable insight by analyzing thermoregulatory strategies in the context of an evolutionary game. The simplest of games occurs between thermoregulators of a single species, which would be described as a game between competitors. This simple model could then be generalized to explore the properties of more complex games, such as the evolutionary interaction between a group of thermoregulators and their predators. In either case, think of behavioral thermoregulation as a process

THERMAL GAMES

of choosing among potential microhabitats, which differ not only in their operative temperatures but also in their densities of food, competitors, and predators. Adopting this point of view enables us to draw on the rich body of theory that was developed for analyzing the optimal use of habitats in an evolutionary game. Game models of habitat choice have steadily accumulated since the early 1970s (reviewed by Sih 1998). All of these models describe a heterogeneous environment as a mixture of two or more patches that differ in resources; here, we shall explicitly consider thermal resources. In the simplest case, we can envision a matrix of warm and cool patches. If we were ambitious, we might extend this analysis to a continuum of thermal patches occurring in equal or unequal frequencies. We might also envision a scenario in which movements among patches impose costs. Here, we shall explore the properties of very simple models to see how novel hypotheses emerge from the study of thermal games.

8.3.1 Competition during thermoregulation To see how a game between competitors can inﬂuence thermoregulatory behavior, we shall adapt a simple model of habitat choice described by Sih (1998). Imagine an environment consisting of two patches that differ in their mean operative temperatures (T1 and T2 ) and their densities of food (F1 and F2 ). In this simple scenario, thermoregulation consists of choosing between these alternative microhabitats on the basis of temperature. Naturally, this choice has repercussions for the performance of an individual. In this model, we can think of performance as the ability to consume and process food; speciﬁcally, performance equals the rate of energy assimilation per unit of food available in a patch. The expected performance of an individual in patch i equals P(Ti ). Therefore, an individual’s rate of energy gain (E) differs between patches because of food and temperature: E(P(T1 ), F1 ) = E(P(T2 ), F2 ). As with other models (e.g., Huey and Slatkin 1976), the optimal strategy of thermoregulation maximizes the net rate of energy gain (assuming an individual expends no energy when moving between patches). Because we desire a thermal game, we must introduce some form of competition among

201

individuals. The actual mechanism of competition does not matter too much; either exploitative competition for food or interference competition for basking sites would yield similar results. In both cases, competition reduces the energy gain of an individual. Mathematically, we can introduce competition by making energy gain a function of the density of competitors as well as food and temperature (Sih 1998): E=

Pi (Ti )Fi , Nix

(8.1)

where Ni equals the density of individuals in patch i. This function enables some ﬂexibility regarding the intensity of competition. When x = 1, individuals compete so severely that a doubling in their density cuts each individual’s performance in half. When x < 1, individuals compete less severely. By setting x = 0, we can also ﬁnd the optimal strategy in the absence of competition. How should competitors distribute themselves between thermal patches? To answer this question, we shall invoke the concept of an ideal free distribution (Fretwell and Lucas 1969). When individuals conform to an ideal free distribution, they occupy patches such that all receive the same rate of energy gain. If any individual changes its behavior, its rate of energy gain would decrease. This loss of energy should lead to further changes in behavior that restore the ideal free distribution. Think of this distribution as the outcome of ﬂexible behaviors shaped by generations of natural selection. To ﬁnd the ideal free distribution, we must set E(P(T1 ), F1 , N1 ) equal to E(P(T2 ), F2 , N2 ) and solve for N1 /N2 . In a population of competitors, the ideal free distribution depends on the disparity of operative temperatures and the intensity of competition: N1 P(T1 )F1 1/x = . (8.2) N2 P(T2 )F2 As we can see from this equation, individuals should distribute themselves according to their relative rate of energy gain in the two patches. During severe competition (x = 1), we should expect twice the number of individuals in the warmer patch if this patch enables twice the rate of energy gain (Fig. 8.1). As competition becomes less severe (i.e., x → 0), the proportion of individuals occupying the warm patch should increase. Nevertheless, we

202

T H E R M A L A D A P TAT I O N

4 Number of competitors

N=1

3 2 Cool

Figure 8.1 In an environment with two patches, the ideal free distribution of competitors should match the distribution of thermal resources. If individuals in the warm patch gain energy twice as fast as individuals in the cool patch, competitors should spend twice as much time in the warm patch.

should expect some individuals to occupy the cooler patch as long as this patch enables some performance. Only in the absence of competition (i.e., x = 0) should all individuals congregate in the warmer patch. But without competition, the model no longer represents a thermal game. The ideal free distribution becomes particularly interesting when food and temperature interact to determine the quality of a patch. Hughes and Grand (2000) explored this phenomenon by modeling the optimal distribution of ﬁsh between two thermal patches. From Chapter 3, recall that food availability affects the thermal sensitivity of growth rate. In ﬁsh, high temperatures maximize growth when food is abundant, but low temperatures maximize growth when food is scarce (see Fig 3.3 and Huey 1982). When many ﬁsh compete for food, this interaction between food and temperature devalues a warm patch beyond the simple effect of competition (Fig. 8.2). At low densities of ﬁsh, all or most ﬁsh should congregate in the warmer patch, which enables rapid growth. As the number of ﬁsh increase in the warm patch, each ﬁsh not only receives less food but also grows less efﬁciently (i.e., grows less per unit of food). At some density, a new ﬁsh entering the warm patch would be better off entering the cool patch instead, because the higher growth efﬁciency in this patch more than compensates for the lower food intake. When many competitors coexist, most ﬁsh will congregate in the cool patch because the limited food available to each ﬁsh causes very poor growth in the warm patch. If the strength of the interaction depends on body size, individuals should segregate between patches according to their

1

Number of individuals

Warm

N=5

4 3 2 1

12

N = 15

8

4

0 6˚C 14˚C Water temperature Figure 8.2 The interaction between food and temperature affects the ideal free distribution of ﬁsh. At low densities (N = 1 or 5), ﬁsh should spend most of their time in the warm patch. Yet, competition at a higher density (N = 15) reduces the feeding rate and growth efﬁciency of ﬁsh in the warm patch, causing ﬁsh to spend more time in the cool patch. The renewal rates of food (5000 mg day−1 ) were equal for both patches. Energetics of growth were based on empirical data for brown trout (Salmo trutta). Based on Hughes and Grand (2000).

size; all else being equal, large ﬁsh should prefer the cooler patch because they grow less than small ﬁsh given the same food intake and body temperature (Hughes 1998). If this interaction represents the rule rather than the exception, the behavior of organisms in nature should deviate from a simple matching of either food or temperature. An elegant experiment conducted by Lampert and colleagues (2003) illustrates the power of game theory to predict thermoregulatory behaviors. These researchers studied the vertical distribution of water

THERMAL GAMES

ﬂeas (Daphnia pulicaria) in a controlled environment. To manipulate distributions of food and temperature precisely, they erected water towers stretching more than 11 m in height! The surface and depths were heated and cooled, respectively, to mimic the thermal stratiﬁcation of a natural lake. The warm epilimnion facilitated growth and development, but most of the food (algae) occurred in the hypolimnion. A model of energetics predicted that the fastest growth would occur at an intermediate depth, just below the thermocline. To achieve an ideal free distribution, water ﬂeas should match their relative potential for growth; hence, the peak abundance of water ﬂeas should occur just below the thermocline. By differentially heating the regions of each tower, the researchers were able to vary the intensity of the thermal stratiﬁcation, creating conditions under which the ideal free distribution would change. Consistent with the prediction of game theory, the distribution of water ﬂeas shifted toward the food-rich hypolimnion as the steepness of the thermocline weakened (Fig. 8.3). In a subsequent experiment, Lampert (2005) showed that another species of Daphnia spreads out toward the foodrich hypolimnion under increasing density. Large individuals altered their distribution the most in response to density, which accords with the model by Hughes and Grand (2000), in which temperature and food interact to determine growth rate.

0

20°/18°C

203

8.3.2 Predation during thermoregulation A more complex and certainly more realistic game occurs when we consider optimal thermoregulation in the presence of predators. Again, we can adapt a simple model by Sih (1998) in which the thermoregulators must balance their use of the superior patch against the risk of predation (see also Hugie and Dill 1994). In mathematical terms, a thermoregulator should maximize the ratio of energy gain to predation risk. Let us assume predators do not concern themselves with behavioral thermoregulation, either because they rely primarily on endothermy or because they perform well over a broad range of temperatures. Thus, the predators care only about maximizing the rate at which they consume their prey (f ). This rate increases with the density of prey in patch i (Ni ), as represented by this linear function: f = cNi ,

(8.3)

where c equals the rate at which predators capture prey. Given ideal free distributions of predators and prey, (1) predators in both patches will gain energy at equal rates and (2) prey in both patches will experience the same ratio of energy gain to predation risk (Box 8.1). Some counterintuitive insights emerge from the ideal free distributions of predators and their prey. If the prey congregate within the superior patch, they become predictable to predators. For this reason,

20°/15°C

20°/10°C

Depth (m)

1 2 3 4 5 6 0

5 10 15 20 25 30 35 Percent

0

5 10 15 20 25 30 35 Percent

0

10

20 30 Percent

40

50

Figure 8.3 The vertical distribution of water ﬂeas (Daphnia pulicaria) in experimental water columns shifted in response to a change in the thermal gradient. In all three cases, the temperature of the epilimnion equaled 20◦ C, while that of the hypolimnion was 18◦ C, 15◦ C, or 10◦ C. The dashed line marks the depth of the thermocline. Open and ﬁlled circles deﬁne the distributions during the day and night, respectively. Error bars represent 1 standard deviation. Adapted from Lampert et al. (2003) with permission from the Royal Society.

204

T H E R M A L A D A P TAT I O N

Box 8.1 Solving for the double ideal free distribution The evolutionarily stable strategy for a system of predators and prey in an environment with two patches must meet two conditions (Sih 1998). First, no individual can increase its performance by switching patches. Second, all individuals must perform equally regardless of which patch they inhabit. For predators, performance equals the rate of energy gain. For prey, performance equals the ratio of energy gain to predation risk. Traditionally, biologists refer to the joint distribution of predators and prey at the evolutionarily stable strategy as a double ideal free distribution. The ideal free distribution of prey equalizes the feeding rates of predators in the two patches: c1 N1 = c2 N2 ,

(8.4)

where ci equals the rate at which predators capture prey. If we assume c1 = c2 , this equation simpliﬁes to yield N1 = N2 .

(8.5)

From eqn 8.4, we learn that a uniform distribution of prey equalizes the feeding rates of predators.

the prey should not prefer either patch, regardless of the thermal beneﬁts. Instead, prey should divide themselves evenly between the patches (N1 = N2 ). Surprisingly, the predators should distribute themselves according to operative temperatures, even though the predators gain no thermal beneﬁts! The ideal free distribution of predators resembles that of the prey in the absence of predators (see eqn 8.2): ρ1 P(T1 )F1 = , ρ2 P(T2 )F2

Likewise, the ideal free distribution of predators must equalize the prey’s ratio of energy gain to predation risk in the two patches: P (T1 )F1 /N1x c1 ρ1

=

P (T2 )F2 /N2x

(8.6)

c 2 ρ2

where ρi equals the density of predators in patch i . Because N1 = N2 at the equilibrium (and c1 = c2 ), eqn 8.5 reduces to ρ1 P (T1 )F1 = . ρ2 P (T2 )F2

(8.7)

Hence, the evolutionarily stable strategy for this system occurs when the predators match the distribution of their prey’s resources while the prey spread evenly throughout the environment. These distributions are stable because any movement of prey from patch 1 to patch 2 would cause predators to shift to patch 2 as well. This response would increase the risk of predation in patch 2 and favor a redistribution of prey. Ultimately, the system would settle back to equal densities of prey in the two patches.

Number of predators

Number of prey

(8.8)

where ρ i equals the density of predators in patch i. Sih (1998) referred to this phenomenon as the leapfrog effect because the distribution of predators matches the distribution of their prey’s resources, instead of the distribution of prey (Fig. 8.4). This surprising result occurs because (1) predators can obtain equal ﬁtness in both patches only when prey are equally abundant in both patches and (2) prey can obtain equal ﬁtness in both patches only when a greater predation risk offsets the greater energy

Warm

Cool

Figure 8.4 In an environment with two patches, the ideal free distributions of prey and their predators become counterintuitive. If prey in the warm patch gain energy twice as fast as prey in the cool patch, predators should spend twice as much time in the warm patch whereas prey should spread evenly between the patches.

gain in the superior patch. The leap-frog effect disappears if we relax the assumption that predators do not compete for prey. To introduce competition, we

THERMAL GAMES

can make the feeding rate of predators a function of their density: f=

cNi . ρi

(8.9)

In this case, both predators and prey should distribute themselves according to the prey’s resources (Sih 1998). Finally, we can imagine that one of the patches provides some refuge from predation. For instance, this patch could be a burrow or some other shelter. We can model the effect of the refuge on predation risk by introducing a coefﬁcient to eqn 8.9 (Sih 1998): cNi f = ki , (8.10) ρi where k1 = 1 makes patch 1 risky and k2 < 1 makes patch 2 safer. Realistically, the operative temperature and food density of the risky patch would exceed those of the refuge (i.e., T1 > T2 and F1 > F2 ). Assuming both competitors and predators compete, the ideal free distribution of prey becomes N1 P(T1 )F1 , (8.11) = k2 N2 P(T2 )F2 and the ideal free distribution of predators becomes P(T1 )F1 ρ1 . (8.12) = ρ2 P(T2 )F2 Therefore, prey should underutilize the warmer patch to take advantage of the refuge. Still, the majority of predators should occupy the warmer patch, even though their feeding rate does not depend on temperature and fewer prey occupy this patch. This result seems counterintuitive but makes sense after some reﬂection. If fewer predators were to occupy the warmer patch, then prey would shift to this patch and enjoy an increase in energy gain. This shift in prey would draw predators back into the warm patch, causing some prey to leave this patch. Thus, the majority of predators must occupy the warm patch to achieve an evolutionarily stable equilibrium.

8.3.3 Relaxing assumptions of simple models The models described in the previous section could be generalized in two ways. First, one could relax assumptions about the biotic environment. For example, additional trophic levels will change the

205

optimal behavior for thermoregulators and their predators, because the predators must then worry about their own mortality. Second, we could relax assumptions about the abiotic environment. For example, the course dissection of an environment into two patches represents an implicit spatial landscape. A more realistic model would account for the spatial relationship among patches and the cost of travel within the environment. Before turning to other kinds of thermal games, let us brieﬂy consider the evolutionary predictions of more complex models of habitat choice. In a community with more than two species, a behavioral cascade can emerge from a thermal game. A behavioral cascade involves an indirect effect of a predator on the behavior of its prey’s prey. For example, consider the simple system in which large ﬁsh feed on small ﬁsh, which in turn feed on zooplankton (Romare and Hansson 2003). In the absence of large ﬁsh, we have a predator–prey interaction that might loosely conform to one of the models described above. Speciﬁcally, we could characterize the environment by two kinds of patches: warm, shallow water and cool, deep water. Assuming equal densities of food in these patches, zooplankton should spread evenly between shallow and deep waters while small ﬁsh tend to occupy the shallow water (see Box 8.1). But what happens when we add a secondary predator to the system? Then, the evolutionarily stable strategy becomes a triple ideal free distribution. With no competition among predators, the triple ideal free distribution resembles the double ideal free distribution with a twist; small ﬁsh should spread evenly between shallow and deep waters while large ﬁsh and zooplankton tend to occupy the shallow water (Rosenheim 2004). In other words, thermoregulators tend to occupy the patch containing superior resources, primary predators spread their risk between patches, and secondary predators tend to occupy the patch containing superior resources for the thermoregulators (Fig. 8.5a). Thus, the prediction of this model appears similar to but fundamentally differs from the leap-frog effect (see Fig. 8.4), particularly with respect to the behavior of the lower trophic levels. Adding yet another trophic level to the system restores the leap-frog effect (Fig. 8.5b). Thus, the conﬂict between exploiting thermal resources

206

T H E R M A L A D A P TAT I O N

(b)

Number of tertiary predators (a) Number of secondary predators

Number of secondary predators

Number of primary predators

Number of primary predators

Number of prey

Number of prey

Warm

Cool

Warm

Cool

Figure 8.5 Adding a secondary predator causes a shift in the optimal use of patches by the primary predator and its prey—a process referred to as a behavioral cascade. As in Fig. 8.4, assume prey in the warm patch gain energy twice as fast as prey in the cool patch. (a) In the presence of secondary predators, primary predators spread evenly between the patches, while their prey use the patches in proportion to the relative energetic beneﬁts. (b) In the presence of secondary predators and tertiary predators, primary predators use the patches in proportion to the relative energetic beneﬁts for prey, while prey spread evenly between the patches.

and avoiding predation risk exists for organisms in complex food chains, as well as simple ones. A more realistic model also results from an explicit description of the spatial environment. Alonzo and colleagues (2003) modeled the optimal movements of penguins and krill among waters of a continental shelf as a consequence of their trophic interactions and thermal physiologies. During this period, penguins leave their breeding grounds periodically to forage at sea. Figure 8.6 depicts the habitats of penguins and krill as characterized by the model. Penguins can move horizontally between inshore or offshore waters and vertically among depths, whereas krill can only move vertically. At deeper depths, krill escape penguins more easily but encounter less food and experience lower

temperatures. These low temperatures reduce rates of energy assimilation and expenditure. Krill also escape penguins more easily during the night than they do during the day. Consequently, diel shifts among depths enable krill to exploit their resources while minimizing predation risk; however, these movements impose an energetic cost. How does the distribution of water temperatures affect the optimal behavior of these organisms? The vertical cline in water temperature directly affects the optimal behavior of krill and thus indirectly affects the optimal behavior penguins. The optimal distributions of the two species depend on whether penguins seek to maximize their daily energy gain or minimize their daily foraging time. If penguins maximize their daily energy gain,

THERMAL GAMES

Penguin breeding grounds

Inshore

207

Offshore

Surface

Surface

Shallow

Shallow

Deep

Figure 8.6 Alonzo and colleagues modeled the optimal use of microhabitats by krill and penguins. Penguins could forage inshore or offshore at any depth, whereas krill could move only among depths. Reproduced from Alonzo et al. (2003) with permission from the Ecological Society of America.

individuals should predominately forage inshore, with some individuals moving offshore at night. The optimal behavior of krill depends on their size, as penguins prefer to eat large krill. During the day, large krill should mainly occupy safe habitats: shallow inshore and deep offshore waters. At night, these krill should move to the surface to feed. Small krill should remain at the surface while medium krill should retreat only to intermediate depths. If penguins minimize the foraging time required to meet their metabolic demands, all penguins should forage inshore (assuming a sufﬁcient density of krill exists there). In this case, krill in offshore waters should remain at the surface. When surface waters are much warmer than shallow and deep waters, more krill spend the day in cool, deep waters to avoid predation and minimize energy expenditure. Thus, warm surface water directly enhances the growth of krill but indirectly reduces the growth of penguins. This model illustrates the complex behaviors that result from a modest increment in the spatial structure of the thermal environment.

8.4 Optimal performance curves in an evolutionary game In Chapter 3, we examined the factors that drive the adaptation of performance curves in heterogeneous environments. In doing so, we concerned

ourselves mainly with the nature of thermal heterogeneity, the ﬁtness consequences of performance, and the genetic variance of thermal sensitivity. Now, let us expand this focus to include a critical factor: the interactions within and among species. As with thermoregulatory decisions, the evolution of thermal sensitivity depends strongly on interactions with competitors, predators, and prey. The predictions emerging from game theory differ qualitatively from those discussed in Chapter 3.

8.4.1 The coevolution of thermal optima between species In a pair of papers, Gavrilets and Kopp developed a game theory of adaptation that applies very nicely to our concept of a performance curve (Gavrilets 1997; Kopp and Gavrilets 2006). Here, we shall focus on the basic components of their theory and the qualitative predictions for the coevolution of two species. Assume species x possesses a performance curve that we can approximate as a Gaussian function (see Chapters 3, 5, and 7 for models based on the same assumption). This particular function has convenient properties for modeling natural selection: as the thermal optimum for performance deviates from the mean temperature of the organism, ﬁtness

208

Te

as follows:

wx = exp −sx (zx − φx )2

exp −γx (zx − zy )2 · fy (zy ) ·

Performance

(a)

T H E R M A L A D A P TAT I O N

(8.13a)

zy

and Temperature (b)

wy = exp −sy (zy − φy )2

exp −γy (zy − zx )2 · fx (zx ). ·

(8.13b)

Fitness

zx

Thermal optimum for performance

Figure 8.7 To apply existing models to thermal games, we must approximate performance curves with a Gaussian function. (a) Deviations between the thermal optimum for performance (Topt ) and the operative environmental temperature (Te ) lead to decreases in the expected performance of individuals in their environment. The solid lines represent performance curves for four genotypes. The gray bar highlights the mean operative temperature. The dashed arrows and lines associated with each performance curve quantify the loss of performance caused by the mismatch between the operative temperature and the thermal optimum. The gain in performance as the thermal optimum approaches the operative temperature mirrors the Gaussian form of the performance curve. (b) The phenomenon described in (a) leads to stabilizing selection of the thermal optimum.

declines according to a Gaussian function that mirrors the performance curve (Fig. 8.7). In other words, a Gaussian performance curve leads to stabilizing selection of the thermal optimum. If the thermal sensitivity of performance were the only selective pressure, the thermal optimum would evolve to match the mean temperature experienced by the species (see Chapter 3). However, species x participates in a thermal game with a second species, y. The interaction between species x and species y imposes directional selection of their thermal optima. The contributions of stabilizing and directional selection to the ﬁtness of each species (wi ) can be characterized

The left-hand term of each function represents stabilizing selection imposed by the thermal sensitivity of performance; for species i, zi equals the thermal optimum, φi equals the mean body temperature, and si scales inversely proportional to the performance breadth. The righthand term represents directional selection imposed by the interaction between species; γi equals the effect of the interaction on species i, and fi (zi ) equals the frequency of individuals with phenotypes zi . Because the signs of γx and γy determine the nature of the interaction, we can use this general model to understand the dynamics of several thermal games. When both γx and γy are positive, the species engage in a mutualistic interaction. When both γx and γy are negative, the species engage in a competitive interaction. When γx is positive but γy is negative, species x and y are predator and prey (or parasite and host), respectively. Regardless of the ecological interaction, the strength of directional selection depends on the difference between the thermal optima of the two species. Assuming individuals choose microhabitats that maximize their ﬁtness, the deviation between thermal optima determines the degree to which the species can (or must) segregate over space and time. Such segregation between competitors (Cerda et al. 1998) or between predators and prey (Joern et al. 2006) occurs in natural environments (but see Kronfeld-Schor and Dayan 2003). Two different approaches were used to model genetic constraints during coevolution. In the ﬁrst approach (Gavrilets 1997), the thermal optimum represents a quantitative genetic trait, whose

THERMAL GAMES

209

Box 8.2 Modeling the genetics of a performance curve Kopp and Gavrilets (2006) supplemented their analysis of the weak-selection approximation with an analysis of coevolution under more realistic genetic constraints. In these simulations, the continuous traits that we are calling thermal optima (x and y ) were affected by L loci, each having two alleles. Alleles at these loci affect the phenotype in an additive fashion. Alleles at each locus were coded as 1 or 0, yielding phenotypic effects of +α/2 or −α/2, respectively. Values of α were drawn from a Gaussian distribution with a mean of zero; therefore, most alleles had very small positive or negative effects on the thermal optimum while a few had large effects. By deﬁning the mean phenotypic value of x and y as mx and my , respectively, they could calculate the range of phenotypes: mx −

L i =1

αi < zx < mx +

L

αi

(8.14a)

i =1

evolution depends on a genetic variance. In the second approach (Kopp and Gavrilets 2006), the contribution of alleles at many loci was modeled explicitly. Because allelic models require far more analytical complexity, the modelers had to make some simplifying assumptions to explore the coevolutionary dynamics. Analytical results were computed for the special case of weak selection and no genetic drift or mutation (the weak selection approximation). More general scenarios demanded computer simulations to deﬁne the model’s behavior (Box 8.2; see also Kopp and Gavrilets 2006; Kopp and Hermisson 2006). The course of coevolution depends on the nature and strength of the biotic interaction, but most scenarios lead to predictions that differ qualitatively from those of traditional optimality models. For the weak selection approximation, the difference between the thermal optima of the species at the equilibrium is given by: z¯ x − z¯ y 1 = . (8.15) φx − φy 1 + γx /sx + γy /sy Although this equation cannot be deﬁned when the two species experience the same mean body temperatures (i.e., when Nx = Ny ), this condition should never exist because each species will have a unique

and

my −

L i =1

αi < zy < my +

L

αi .

(8.14b)

i =1

Evolutionary dynamics were modeled by assuming discrete, non-overlapping generations. In each generation, the populations underwent selection, random mating, recombination, and mutation. Mutation converted an allele coded as 0 to an allele coded as 1, or vice versa. Given the complexity of this model, Kopp and Gavrilets resorted to computer simulations to ﬁnd evolutionarily stable strategies. Fortunately, the behavior of this complex allelic model supported the conclusions drawn from a quantitative genetic model (Gavrilets 1997).

operative temperature in the same microhabitat (see Chapter 2). Using eqn 8.15, we can predict the thermal optima of two species engaged in a competitive, mutualistic, or an exploitative interaction. Recall that γx and γy determine the nature of the interaction between the species. For competitors, both γx and γy are negative, which causes their thermal optima to diverge (Fig. 8.8a); consequently, neither species would exhibit maximal performance at the mean operative temperature of its environment. For mutualists, both γx and γy are positive, which causes the thermal optima of the two species to converge (Fig. 8.8b). For predator and prey, γx is negative (prey) and γy is positive (predator), which could lead to one of several outcomes depending on the relative strengths of directional and stabilizing selection in the two species (Fig. 8.8b–f). When prey experience strong stabilizing selection, the predators converge on the same thermal optimum as the prey. When prey experience strong directional selection or weak stabilizing selection, the prey escape predation by evolving an extreme thermal optimum. When the genetic variance of the prey limits their escape from the predators, disruptive selection produces two distinct thermal optima in the prey and an intermediate thermal optimum in the predators. Finally, the

210

T H E R M A L A D A P TAT I O N

Thermal optimum of species y

Competition (a)

Predator–prey (d)

Mutualism/ predator–prey (b)

Predator–prey (e)

Figure 8.8 The coevolution of thermal optima between two species depends on the nature of their interaction, their breadths of performance, and the additive genetic variances. (a) The thermal optima of competitors should diverge from the mean operative temperature to reduce the intensity of competition. (b) Stabilizing selection should cause the thermal optima of mutualists to converge on the same value. (c) Disruptive selection can cause distinct thermal optima to evolve within a population of prey, while an intermediate thermal optimum evolves in the population of predators. (d and e) The thermal optima of predators and prey can cycle repeatedly as both experience selection in one direction imposed by their interaction, followed by selection in the opposite direction imposed by the thermal environment. (f) Prey can evolve an extreme thermal optimum while stabilizing selection prevents the predators from matching the thermal optimum of the prey. Adapted from Kopp and Gavrilets (2006) with permission from Blackwell Publishing.

Predator–prey (c)

Predator–prey (f)

Thermal optimum of species x

thermal optima of predators and prey cycle indefinitely when the prey have a stronger incentive to win (|γx | > γy ) or have greater genetic variance and experience stronger stabilizing selection (sx > sy ).

8.4.2 The coevolution of thermal breadths between species Although frequency-dependent selection clearly affects the evolution of thermal optima, performance breadths can also coevolve. Ackermann and Doebeli (2004) developed a model that offers stimulating insights on the evolution of performance breadths in a thermal game. The model focuses on genotypes that compete for a continuously distributed resource, which we shall think of as temperature. Because the genotypes reproduce asexually, they could represent genetic diversity within a single species or between two competing species. The performance curve was modeled as a variant of the Gaussian function:

e−cz2 P(Tb ) = exp −(z1 − Tb )2 /2z22 , (8.16) 2π · z2 where Tb equals the temperature of the organism, and z1 and z2 deﬁne the thermal optimum and performance breadth. This function strongly resembles the one used to model the evolution of thermal sensitivity in Chapter 3, with the distinction that a new parameter (c) deﬁnes the beneﬁt of specialization. If c = 0, the performance

curve evolves according to the typical constraint imposed by specialist–generalist tradeoffs: the integral of the performance curve remains constant. If c > 0, specialist–generalist tradeoffs become more severe, such that widening the performance breadth reduces the integral of the curve. If c < 0, the reverse becomes true, and generalists have an advantage over specialists. Because the model considers a single organismal performance affected by a single environmental resource, the constraints imposed when c = 0 cannot be interpreted as acquisition or allocation tradeoffs (see Box 3.5). Instead, these constraints must stem from some unspeciﬁed physical process associated with specialization or generalization. Whether or not we believe such constraints exist will become an important consideration once we have examined the behavior of the model. To apply this model, we must assume individuals occupy thermal niches according to their performance curve.14 Consequently, most individuals of a species will occupy microhabitats that provide them with their optimal temperature, whereas fewer individuals will occupy microhabitats that provide a higher or lower temperature. No individual can occupy a microhabitat in which the operative temperature falls outside its performance curve. In the absence of any organisms, a Gaussian function 14

The distribution of individuals assumed by this model equals the ideal free distribution when individuals of the same species compete for resources.

THERMAL GAMES

Frequency

(a) Frequency distribution of temperatures

Operative temperature

Performance

(b) Performance curve of resident genotype

z1

z2 Body temperature

Frequency

(c) Frequency distribution of temperatures available to an invading genotype

z1

Operative temperature Figure 8.9 Thermal resources can be exploited by a resident genotype, which depletes the availability of operative temperatures (and associated resources) for other genotypes. In this example, the resident genotype uses thermal resources (a) according to its thermal performance curve (b), causing the greatest reduction in thermal resources near the resident’s thermal optimum (c). The parameters z1 and z2 determine the thermal optimum and performance breadth, respectively. Adapted from Ackermann and Doebeli (2004) with permission from Blackwell Publishing.

deﬁnes the availability of operative temperatures (Fig. 8.9a). These temperatures are distributed in space such that any individual can access any microhabitat without paying a cost. Nevertheless, the presence of an individual in a microhabitat prevents another individual from effectively using this site, thus altering the frequency of available temperatures (Fig. 8.9b and c). To be clear, Ackermann and Doebeli envisioned the environmental axis as a biotic resource, which grew logistically in response to consumption, but we can apply their model to an

211

abiotic resource without too much imagination. We need only assume that individuals occupy microhabitats, thereby depleting the availability of certain operative temperatures to other individuals. Both the coexistence of competing genotypes and the evolution of their performance curves depend on the strength of the constraint imposed by specialist– generalist tradeoffs (Fig. 8.10). When these tradeoffs impose a weak or modest constraint (c ≤ 0), a common genotype evolves. This genotype has a thermal optimum equal to the mean operative temperature and a performance breadth just wide enough to exclude all other genotypes. This performance breadth greatly exceeds the one predicted by a model of frequency-independent selection (Gilchrist 1995; see Chapter 3). When specialist–generalist tradeoffs impose a strong constraint (c > 0), distinct genotypes evolve by disruptive selection. As the cost of generalization increases, narrower performance breadths should evolve and more genotypes should coexist (compare Fig. 8.10b and Fig. 8.10c). Under this severe constraint, an important prediction of the model emerges: no phenotype should have a thermal optimum equal to the mean operative temperature.

8.4.3 Gene ﬂow and the coevolution of thermal optima From Chapter 3, recall that gene ﬂow among disparate thermal environments can either facilitate or impede the adaptation of performance curves (see Section 3.9). Here, we shall use a model developed by Day (2000) to explore a thermal game between two populations that experience disparate thermal environments but exchange genes via migration. Imagine two environments that both have a Gaussian distribution of operative temperatures but have different mean temperatures. Each generation, individuals migrate between environments in proportion to the sizes of their populations. In the absence of migration, the divergence of thermal optima within environments would reduce the impact of competition. Would migration between populations enhance or impede this disruptive selection? The answer depends on the symmetry of performance between environments. Environmental symmetry means that locally adapted organisms in each

212

T H E R M A L A D A P TAT I O N

Thermal optimum

(a) Classic tradeoff

(c) Ultra-severe tradeoff

3

3

0

0

0

–3

–3 0

Thermal breadth

(b) Severe tradeoff

3

5000

10 000

–3 0

10 000

20 000

0

3

3

3

2

2

2

1

1

1

0

0 0

5000

10 000

Time (generations)

10 000

20 000

10 000

20 000

0 0

10 000

20 000

Time (generations)

0

Time (generations)

Figure 8.10 Computer simulations show that the evolution of thermal performance breadth depends on the constraint imposed by specialist–generalist tradeoffs. (a) The typical constraint, in which the integral of the performance curve remains constant (c = 0), leads to the evolution of a wide performance breadth. (b and c) More severe tradeoffs cause the evolution of distinct thermal optima and relatively narrow performance breadths. Values of c were 0.6 and 1.25 for simulations shown in (b) and (c), respectively. Thermal optima are plotted relative to the mean operative temperature. Adapted from Ackermann and Doebeli (2004) with permission from Blackwell Publishing.

environment achieve the same rates of energy gain, whereas environmental asymmetry means that local adaptation enables a greater rate of energy gain in one of the two environments. Asymmetry of performance can result from either of two mechanisms. First, one environment may contain more food, enabling individuals to gain energy faster than they could in the other environment. Second, a warm environment might enable greater performance because of thermodynamic constraints (i.e., hotter is better; see Chapter 4). When environmental symmetry exists, migration between environments favors a single thermal optimum that lies in between the mean operative temperatures of the two environments. The critical rate of migration depends on the mean performance in each environment; if locally adapted individuals perform very well, migration must occur more frequently to favor an intermediate thermal optimum. Thus, greater productivity or warmer conditions tend to favor disruptive selection of the thermal optimum. When environmental asymmetry exists, stabilizing selection should produce a single thermal optimum. The value of this thermal optimum depends on the frequency of migration; infrequent migration favors a thermal optimum closer to the mean temperature of the better environment, whereas frequent migration favors a thermal optimum closer to the mean

temperature of the poorer environment. In general, thermal heterogeneity coupled with gene ﬂow inhibits the divergence of thermal optima between competitors.

8.5 Life-history evolution in a thermal game The concept of a thermal game applies not only to the evolution of thermoregulation and thermosensitivity but also to the evolution of the life history. To illustrate this point, let us consider how the predictions of a simple model of optimal reproductive allocation depend on the frequency dependence of selection. From Chapter 6, recall that we used a model by McGinley and colleagues (1987) to determine the optimal size(s) of offspring in a patchy environment. For example, assume warm patches enhance the performance of offspring, whereas cold patches impair their performance. When parents can direct their offspring to a warm patch, the model predicts that mothers should produce relatively small offspring. But when the performance of offspring depends on their density, not all parents can reap the beneﬁts of a warm patch. If enough parents were to direct their offspring to warm patches, the quality of these patches would fall below that

THERMAL GAMES

of the cold patches. At that point, other parents would do better to direct their offspring to cold patches. Obviously, each parent should allocate its resources based on the environment they expect their offspring to experience (see Chapter 6). But the quality of the environment depends on the reproductive decisions of other mothers. Thus, the optimal strategy becomes the evolutionarily stable strategy. The addition of frequency-dependent selection to the model of optimal offspring size revealed a new condition under which variation in offspring size might evolve. Through computer simulations, McGinley and colleagues (1987) calculated a range of evolutionarily stable strategies, varying the size of small offspring, the ability to direct offspring to the correct habitat, the intensity of density dependence, and the difference in quality between habitats. Based on their simulations, we can draw several conclusions about reproductive decisions in a thermal game. Under random dispersal, the evolutionarily stable strategy always results in a single offspring size: either the optimal size for a warm patch or the optimal size for a cold patch. Even when parents can direct their offspring perfectly to a warm or cold patch, the evolutionarily stable strategy often involves a single offspring size. A mixture of offspring sizes is favored only when very strong density dependence exists, such that a rare genotype directing large offspring to a cold patch fares better than a common genotype directing small offspring to a warm patch. The critical level of density dependence required to favor a mixture of offspring sizes increases as the quality of the patches becomes

213

more disparate. Therefore, strong density dependence coupled with moderate thermal heterogeneity favors a mixed reproductive strategy.

8.6 Conclusions Although the models summarized in this chapter are far too simple to be realistic, we should still come away with a sobering conclusion: a thermal game can generate patterns of adaptation that differ drastically from the patterns generated by traditional optimality models. Natural selection might favor a very low accuracy of thermoregulation when an organism must deal with competitors or predators. The same ecological interactions can shape the evolution of performance curves in surprising ways. Competition causes the thermal optimum to deviate from the mean operative temperature of the environment. Predation can cause endless cycles of adaptation in which the thermal optima of predators and prey rarely match their mean operative temperatures. Equally novel insights will likely emerge from the analysis of other thermal games. For example, the decision to mature early or late should depend on the number of conspeciﬁcs who make a similar decision. What effect will competition or predation have on the evolution of this life-history decision? How might these ecological interactions inﬂuence the coadaptation of behavior, physiology, and life history within species? Given what we know already, I am eager to learn the full extent to which game theory will change our perspective of thermal adaptation.

CHAPTER 9

Adaptation to Anthropogenic Climate Change

As I type these words, former Vice President Al Gore and members of the Intergovernmental Panel on Climate Change (IPCC) head to Stockholm to collect a Nobel Peace Prize. The prize was awarded “for their efforts to build up and disseminate greater knowledge about man-made climate change, and to lay the foundations for the measures that are needed to counteract such change.” These efforts include multiple syntheses of scientiﬁc data and the recent documentary An Inconvenient Truth—all of which brought the issue of climate change into the public spotlight. While politicians, reporters, and laypeople continue to question the causes of the recent global warming, few scientists doubt the inﬂuence of human activities. The mean global temperature increased by 0.6◦ C during the last century, and reasonable models predict an additional increase of 2–4◦ C by the end of this century (IPCC 2007). Such scenarios could have startling consequences for ecosystems and biodiversity. But make no mistake. Adaptation will play some role in determining these consequences. Throughout this book, we have explored the ways in which populations can adapt to thermal heterogeneity. Current patterns of phenotypic variation were produced by evolution during countless generations of thermal change. The direction and rate of change have certainly varied regionally over space and globally over time. But at this moment, many ecosystems are warming faster than they have for thousands of years (IPCC 2007). The biological impacts of this environmental perturbation are already under way. Indeed, some would say we are in the midst of a mass extinction caused by the advance of one species: Homo

214

sapiens. Understanding the role that global warming will play in this drama has become one the greatest challenges facing ecologists. In this ﬁnal chapter, we shall see how ecologists currently model the impacts of climate change and consider how evolutionary biologists can help to improve these models.

9.1 Recent patterns of climate change 9.1.1 Global change According to the IPCC (2007), our planet warmed by 0.6◦ C during the last century and continues to warm in the present century. The case for global warming is open and shut. Of the last 12 years, 11 were among the 12 warmest years in recorded history (1850– 2006). Diel variation has become smaller because nighttime temperatures have increased more than daytime temperatures. Seasonal variation has also become smaller because winter temperatures have increased more than summer temperatures. Heat waves have become more frequent, and cold snaps have become less frequent. The indirect effects should startle even a layperson. Some regions, such as the Mediterranean, southern Africa, and southern Asia, receive less precipitation than in previous decades. Other regions receive more precipitation concentrated into fewer events, leading to greater runoff. The melting of ice in the Arctic has contributed to an increase in sea levels. To make matters worse, global warming has accelerated over time; the rate of warming during the last decade exceeded the rate during the last 50 years! Identifying the primary cause of global warming has enabled scientists to make projections about the

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

future. Few scientists doubt the mechanistic link between the accelerating emissions of greenhouse gases and recent patterns of environmental warming (IPCC 2007). Greenhouse gases (e.g., carbon dioxide) inﬂuence the absorption, scattering, and emission of radiation, leading to an increase in surface and air temperatures. Because these emissions stem from the industries that supply energy, transportation, and other products, we can expect current trends to continue unless we reduce or cease these activities. Even if we stopped emitting greenhouse gases immediately, the globe would likely warm by 0.5◦ C during the next century (IPCC 2007). A more realistic scenario includes an increase in the annual emission of carbon dioxide by 40–110% between 2000 and 2030, causing the earth to warm 0.6◦ C during this period alone (IPCC 2000, 2007). If emissions continue to accelerate, the mean surface temperature of the earth might increase by as much as 6◦ C by 2100 (IPCC 2007). Although predictions on the scales of decades to centuries carry great uncertainty, the prediction offered by the IPCC in 1990 accords well with the subsequent rate of warming (predicted

–1 to –0.7

–0.7 to –0.3

> –0.3

215

rate, 0.15–0.3◦ C per decade; observed rate, 0.2◦ C per decade).

9.1.2 Regional change Although the pattern of global warming appears relatively straightforward, regional changes in temperature have been far more complex. The most rapid warming has occurred at high northern latitudes, with the Arctic having warmed twice as fast as the global average. Not surprisingly, the continents have warmed more rapidly than the oceans. Among oceanic environments, warming has occurred most slowly in the Southern Ocean surrounding Antarctica. Even these patterns obscure heterogeneity of thermal change on smaller scales (Walther et al. 2002). A detailed look reveals that some regions have warmed by as much as 1◦ C per decade while others have cooled at a similar rate (Fig. 9.1). Clearly, these regional patterns matter more to organisms and populations than do global patterns.

< –0.3

0.3 to 0.7

0.7 to 1

Figure 9.1 Environmental temperatures have changed at different rates throughout the world. This map shows rates of change (◦ C per decade) between 1976 and 2000, relative to normal temperatures for 1961–1990 (see Box 2.1). Reprinted by permission from Macmillan Publishers Ltd: Nature (Walther et al. 2002), © 2002. See plate 3.

216

T H E R M A L A D A P TAT I O N

9.1.3 Local change On local scales, the activities of humans have added to the thermal change caused by greenhouse gases. Both the destruction of natural habitats and construction of human dwellings have generated marked changes in the thermal landscape. Devegetation and urbanization have raised local temperatures of both terrestrial and aquatic habitats (Grimm et al. 2008; Nelson and Palmer 2007; Roth 2002). Geographers refer to the elevations of air and surface temperatures in cities as urban heat islands. The magnitude of this thermal effect scales logarithmically with the population of a city (Oke 1973). In very large cities, urban warming has matched or exceeded the rate of global warming. Urban heat islands possess complex spatial and temporal structures (Arnﬁeld 2003; Weng et al. 2004). Spatial variation depends on the historical factors that shape the development and growth of each city. Temporal variation appears less idiosyncratic; within a particular city, urban temperatures exceed rural temperatures more during the summer than any other time of year.15 Because urban warming and global warming have proceeded independently (Parker 2004), the two processes combine to magnify potential impacts on organisms.

9.2 Observed responses to recent thermal change In the past decade, scientists have made considerable progress toward assessing the biological consequences of recent climate change. A mixture of detailed case studies and comprehensive metaanalyses uncovered four major trends: (1) shifts in phenology, (2) shifts in geographic ranges, (3) disruption of ecological interactions, and (4) changes in primary productivity (see reviews by Hughes 2000; Parmesan 2006; Parmesan and Galbraith 2004; Walther et al. 2002). Although these trends could stem from factors other than anthropogenic climate 15

The time of day at which urban heat islands reach their greatest intensity depends on whether one characterizes these heat islands by surface or air temperatures. The difference between urban and rural air temperatures peaks during the night, whereas the difference between urban and rural surface temperatures peaks during the day (reviewed by Arnﬁeld 2003; Weng et al. 2004).

change (Forchhammer and Post 2000), some effort has been made to control for other factors in these analyses. Below, we shall examine these four trends in some detail and consider their potential impact on the future of biological systems.

9.2.1 Shifts in phenology By far, the most widely documented impacts of global warming involve phenology. Because temperature determines the performance of most organisms, environmental warming should enable organisms to initiate activity earlier in the spring and sustain activity later in the fall. Consistent with this expectation, the timing of biological events in spring has advanced considerably during recent decades. Of 385 species of British plants, 16% have advanced their date of ﬂowering, whereas only 3% have delayed this date (Fitter and Fitter 2002). A global meta-analysis of 143 studies of amphibians, birds, invertebrates, trees, and other plants documented a signiﬁcant advance of vernal activity (Root et al. 2003). For the affected species, the mean rate of advance was 5.1 days per decade. In a similar analysis of 677 species, 62% of species showed earlier activity in spring, such as breeding, nesting, ﬂowering, migrating, or budding (Parmesan and Yohe 2003); on average, spring activity shifted 2.3 days per decade. These two estimates of phenological advance (5.1 days vs. 2.3 days) reﬂect both taxonomic and methodological differences between analyses. By combining the data, Parmesan (2007) estimated a mean advance of 2.8 days per decade, with amphibians advancing the most and plants advancing the least (Fig. 9.2). As we might expect from regional differences in warming, vernal activities have advanced more in some locations than in others. As expected, species at high latitudes have advanced more than species at low latitudes. Similarly, phenology has changed more in urban environments than it has in rural environments (Roetzer et al. 2000; White et al. 2002), revealing the additive effects of urban and global warming. Still, much of the variation in phenological change remains unexplained by these factors; for example, latitude accounted for less than 5% of the variation in the advance of vernal activity (Parmesan 2007).

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

These meta-analyses did not attempt to document temporal variation in the rate of phenological advance, but we should expect such variation for two reasons. First, the rate of warming experienced by particular stages of the life cycle has accelerated in recent decades. Case studies indicate that species respond to local rates of warming rather than the global rate. For example, some birds lay their eggs at times of the year that have been less affected by global warming; this difference in life history might explain why the laying dates of some North American species have advanced while the laying dates of others have not (Torti and Dunn 2005). Second, even a constant rate of warming can cause a variable rate of phenological advance. Because a linear increase in temperature causes a nonlinear increase in performance, the physiological consequences of constant warming will vary over time.

9.2.2 Shifts in geographic ranges In addition to shifting their activity in time, organisms have also shifted their activity in space. The geographic ranges of many species have shifted northward or upward in recent decades. Of 35 species of butterﬂies, 63% have moved their entire range poleward while only 3% have moved southward (Parmesan et al. 1999). Of 434 species whose ranges have shifted historically (17–1000 years), 80% have shifted in accord with predictions based on climate change (Parmesan and Yohe 2003). For 99 species of birds, butterﬂies, and alpine herbs, range boundaries have shifted an average of 6 km northward or 6 m upward per decade (Parmesan and Yohe 2003). By no means have shifts in geographic ranges been conﬁned to terrestrial environments. Over the last 40 years, warm-adapted copepods have moved 10◦ N and have begun to replace coldadapted copepods (Beaugrand et al. 2002). Similarly, warm-adapted species of algae have shifted north, but cold-adapted species have shifted either north or south (Lima et al. 2007a). Either range contraction or global extinction is a likely outcome for many species, particularly when body size, life history, or an ecological interaction constrains dispersal. Some species have already lost the southern or lower portion of their range (see, for example, Franco et al. 2006). Sadly, species at

217

extreme latitudes or altitudes have nowhere to go; hence, many of these species will face extinction (Parmesan 2006). Thomas and colleagues (2004) used several methods to estimate the risk of extinction for more than 1000 species of animals and plants. They assumed a species–area relationship to link the overlap among species to the risk of extinction. First, they predicted future ranges from climate-envelope models (see Section 9.3). Then, they projected extinction risk for different scenarios of climate change. Because thepotentialfordispersalclearlyaffectsthepredicted shiftinrange, twoextremesituationswerecompared: perfect dispersal and no dispersal. For perfect dispersal, a species’ future range equaled its future climate envelope. For no dispersal, a species’ future range was the intersection of its current and future climate envelopes. Given minimal warming (0.8–1.7◦ C), 9–13% of species were predicted to face extinction under the assumption of perfect dispersal; without dispersal, as many as 31% of species might become extinct. With more extreme warming (>2◦ C), these predictions increased to 21–32% with dispersal and 38–52% without dispersal. Extinctions caused by global climate change might even exceed those caused by local habitat destruction (Thomas et al. 2004). More realistically, global and local insults to the environment will interact to heighten the risk of extinction. In particular, the fragmentation of landscapes by humans will impair the dispersal of organisms during climate change (Parmesan and Galbraith 2004), generating an outcome closer to the scenario of no dispersal than the scenario of perfect dispersal.

9.2.3 Disruption of ecological interactions Because shifts in phenologies and ranges have differed among species, some ecological interactions have become uncoupled (Harrington et al. 1999). In a review of 11 interspeciﬁc relationships, Visser and Both (2005) concluded that ﬁve species advanced too little and two species advanced too much compared with their prey. Uncoupling has occurred over space and time. For example, the northern spread of the brown argus butterﬂy has resulted in some release from parasitism (Menéndez et al. 2008). Similarly, the vernal activities of other insects have advanced more rapidly than their botanical

Tr ees Ot he rs

rub Sh

He rb gra s an sse d s

Bu tte rfl ies

Bir

ds

T H E R M A L A D A P TAT I O N

Am ph ibi an s

218

10

Change in timing (days per decade)

5 0 –5 –10 –15 –20 –25 –30 –35 Figure 9.2 Changes in the timing of vernal activity have varied greatly among taxa. Bars represent the rate of advance (−) or delay (+) of vernal activity for individual species. The group labeled “Others” includes two species of ﬁsh, a species of ﬂies, and a species of mammals. Adapted from Parmesan (2007) with permission from Blackwell Publishing.

hosts (Gordo and Sanz 2005) or their avian predators (Visser et al. 2006). Uncoupling of ecological interactions promises to become more severe as the compositions of local communities change (Peterson et al. 2002). Compositional changes seem inevitable because species differ in their ability to migrate and their susceptibility to extinction. The consequences for community dynamics will depend on whether invading species disrupt the interactions between native species or ﬁll the vacuum created by extinction.

9.2.4 Changes in primary productivity The organismal and populational impacts of climate change have altered the structure and function of ecosystems. Biologists care much about the consequences for primary productivity because of its potential feedback on global warming. Greater productivity of plants creates a sink for carbon dioxide, while less productivity creates a source. Both the emission of carbon dioxide and the increase in environmental temperature should enhance the productivity of plants in certain ecosystems (Kirschbaum 2004). High levels of carbon dioxide enable plants to increase photosynthesis without increasing transpiration. Warm soils promote the release of nutrients

to sustain photosynthesis. Among 32 experiments in which ecosystems were warmed by 0.3–6◦ C for 2–9 years, warming increased plant productivity by 19% and soil respiration by 20% (Rustad et al. 2001); the productivity of plants in cold climates increased the most. Consistent with these experiments, recent warming has enhanced the growth of terrestrial forests, leading to greater sequestration of carbon. Nevertheless, this process will not create enough negative feedback to stabilize the global temperature. The growth of terrestrial forests should cease and may even reverse by the middle of this century (IPCC 2007). Furthermore, the increasing frequency of droughts has led to a drop in the primary productivity of other ecosystems (see, for example, Ciais et al. 2005). All things considered, global warming will continue despite the feedback caused by primary productivity.

9.3 Predicting ecological responses to global warming Although we now recognize multiple impacts of global warming, what impacts should we expect to observe in the future? On one hand, we can certainly expect continued shifts in the phenologies and ranges of species. On the other hand, the magnitudes

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

of these shifts have varied greatly among species in recent decades. Thus, we must move beyond simple inductive reasoning to accurately predict the impacts of global warming. In this section, we shall focus on mathematical models that provide insights about the ecological responses to climate change. By contrasting correlative and mechanistic models, I hope to convince you that mechanistic models offer the greater potential to incorporate evolutionary processes.

9.3.1 Correlative versus mechanistic models Efforts to predict the biological impacts of climate change in general, and global warming in particular, have been based on the concept of the fundamental niche (Hutchinson 1957). The fundamental niche depicts the abiotic factors that limit the success of a genotype or species, having one dimension for every environmental variable we can imagine (see review by Wiens and Graham 2005). In practice, biologists use a limited number of variables to characterize these niches. Thermal variables, such as the mean and variance of environmental temperature, have played a prominent role in deﬁning fundamental niches (Lima et al. 2007b; Walther et al. 2007). To see how climate change might affect a species, researchers model the fundamental niche of a species and then identify the location of this niche at some time in the future. In recent years, biologists have used both correlative and mechanistic approaches to deﬁne fundamental niches (Pearson and Dawson 2003). Correlative models deﬁne the niche by relating current geographic distributions to local environmental conditions (see Elith et al. 2006). Assuming a species will track the geographic position of its niche, we can use a correlative model to predict a shift in a species’ distribution during climate change (Hijmans and Graham 2006). Because correlative models require only environmental data associated with known localities (Graham et al. 2004; Guisan and Thuiller 2005), these models provide a convenient approach to predicting global distributions. Correlative models can also deal with species interactions (Araujo and Luoto 2007; Sutherst et al. 2007) and geographic variation (Murphy and Lovett-Doust 2007) to some degree. Nevertheless, these models conﬂate

219

abiotic and biotic factors, fail to account for dispersal, and ignore potential evolutionary responses (Pearson and Dawson 2003; Soberón and Peterson 2005). In sharp contrast, mechanistic models explicitly link the environmental tolerances of individuals to the dynamics of populations. Their mathematical functions not only make the underlying assumptions transparent, they also provide opportunities to relax these assumptions and incorporate further biological detail. For example, we could modify a mechanistic model to account for biotic interactions, dispersal limitations, and evolutionary adaptation. Unfortunately, implementing a mechanistic model requires extensive knowledge of the behavior, physiology, and life history of a species. Do we really need complicated mechanistic models, or can we rely on simpler correlative models to predict the impacts of climate change? No one can deﬁnitively answer this question at the moment, but recent attempts to hindcast responses to climate change have me leaning in one direction. When hindcasting, one predicts the current distribution of a species with a model based on environmental and biological data from the past (see Lima et al. 2007b; Walther et al. 2005). Araujo and colleagues (2005) conducted the most extensive hindcasting effort thus far, predicting the distributional changes of 116 species of birds between 1967 and 1991. For each species, these researchers applied four correlative models and several sets of assumptions. For 90% of the species, the models disagreed as to whether the geographic range should have expanded or contracted during the two decades. To make matters worse, when all models made similar predictions, these predictions were qualitatively wrong for half of the species. Why do correlative models fail to accurately predict responses to climate change? Many factors— including biotic interactions, geographic dispersal, and evolutionary adaptation—interact with climate change to determine the success of a species (Jackson 1974; Pearson and Dawson 2003; Soberón and Peterson 2005). Current distributions appear limited by dispersal in some species and biotic interactions in others. For example, the limpet Collisella scabra grew well when transplanted either 100 or 200 km north of its current range boundary, suggesting barriers to dispersal limit a northern expansion (Gilman

220

T H E R M A L A D A P TAT I O N

2006). In contrast, the distribution of the butterﬂy Aporia crataegi appears constrained by the absence of host plants at high elevations. (Merrill et al. 2008). Recent experiments challenge the notion that we can cleanly separate biotic and abiotic determinants of the niche; for example, the outcomes of competitive interactions (Davis et al. 1998; Jiang and Kulczycki 2004; Jiang and Morin 2004) and host– parasite interactions (Harvell et al. 2002; Poulin 2006) depend on environmental temperature. Even if correlative models were to accurately capture the factors that limit the current distribution, adaptation to environmental change affects the future distribution. These problems associated with correlative models can lead to overprediction or underprediction of a geographic distribution (Pulliam 2000; Soberón and Peterson 2005). A model would overpredict the distribution when individuals cannot disperse to all favorable environments and underpredict the distribution when individuals disperse routinely to unfavorable environments. The same model could overpredict or underpredict the distribution of a species when interactions among species occur. Finally, the model would underpredict the distribution when adaptation occurs. Therefore, the advancement of mechanistic models that account for these processes seems essential for making accurate predictions.

9.3.2 Mechanistic models of responses to environmental warming Mechanistic models differ greatly in their complexity and generality. Some mechanistic models rely on an environmental threshold to deﬁne the fundamental niche, such as thermal constraints on survival, growth, development, or feeding (Cortemeglia and Beitinger 2006; Kearney and Porter 2004; Parker and Andrews 2007; Robinet et al. 2007; Sykes et al. 1996). Other models link these thermal constraints directly to the population dynamics of plants and animals (Dullinger et al. 2004; Edmunds 2005; Pulliam 2000). Here, we shall focus on two models that deal explicitly with population dynamics, because these models could eventually accommodate the potential for thermal adaptation. In the ﬁrst model, environmental temperature determines the development, fecundity, and survivorship of

a thermoconformer (Crozier and Dwyer 2006). In the second model, environmental temperature constrains the duration of activity by a thermoregulator (Buckley 2008). These models offer complementary perspectives of the impacts of environmental warming on the population dynamics and geographic distributions of ectotherms. Crozier and Dwyer (2006) used performance curves to model the population growth and range limits of a thermoconformer in a seasonal environment. They split the life cycle into winter and summer stages. Winter temperatures determine the probability of surviving until summer (S). Summer temperatures determine the net reproduction (R), deﬁned as the number of individuals at the end of the summer that descended from each female alive at the beginning of the summer. The growth rate of a local population (λ) equals the product of winter survivorship and net reproduction: λ = S(Twin ) · R(Tsum ),

(9.1)

where Twin and Tsum equal the mean temperatures of winter and summer, respectively, which are functions of latitude (L) and time (t). Because a population can persist only if λ ≥ 1, Crozier and Dwyer could solve for the northerly limit of the range by calculating the latitude at which λ = 1. To predict the range shift in response to warming, they assumed a simple linear change in winter and summer temperatures (Twin and Tsum ) over latitude and time: Twin = aw − bw L + cw t

(9.2)

Tsum = as − bs L + cs t,

(9.3)

and

where bw and bs determine the rates at which temperatures decrease with increasing latitude, and cw and cs determine the rates at which temperatures increase over time. Under these conditions, the change in the latitudinal limit with time was found: cw R(Tsum ) dTdS + cs S(Twin ) dTdR dL sum win = dt bw R(Tsum ) dTdS + bs S(Twin ) dTdR win

(9.4)

sum

If we assume net reproduction and winter survivorship decline exponentially with decreasing

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

temperatures (i.e., R ∝ eδ·Tsum and S ∝ eρ·Twin ), eqn 9.4 becomes dL δcs + ρcw = , dt δbs + ρbw

(9.5)

where δ and ρ deﬁne the thermal sensitivities of R and S, respectively. Thus, the shift in the range limit depends not only on the rates of summer and winter warming, but also on the latitudinal clines in seasonal temperatures. Crozier and Dwyer used their model to predict shifts in the northern limit of the sachem butterﬂy (Atalopedes campestris) under several warming scenarios. Values of R(Tsum ) and S(Twin ) were estimated from the thermal sensitivities of survivorship and development in butterﬂies from eastern Washington. The model predicted the current limit remarkably well in the western and eastern portions of the USA, but underpredicted this limit in the central portion (Fig. 9.3a). The predicted responses to warming scenarios were complex because winter and summer temperatures interact to alter the fundamental niche over time (Fig. 9.3b). For example, imagine the current range ends at the latitude where the winter temperature falls below 5◦ C. As temperatures rise steadily, the latitudinal boundary for this thermal threshold will shift northward. But the species will not track this boundary faithfully because summer temperatures also affect the range limit (see eqn 9.4). Because many environmental variables limit a species’ distribution, such interactions should become an important focus of mechanistic models. Taking a more complex mechanistic approach, Buckley (2008) modeled the constraints on foraging activity in a thermoregulating ectotherm. She envisioned a territorial, sit-and-wait forager that maintained a preferred body temperature during foraging through energetically inexpensive behaviors. Because the animal could forage only when the environment enabled thermoregulation, warming scenarios would alter the duration of foraging (tf ), and hence the assimilation of energy and the production of offspring. To model thermal constraints in each environment, Buckley calculated the operative temperatures in full sun and full shade; an individual could forage whenever the temperatures

221

bounded by these extremes overlapped with the preferred range of temperatures. When foraging, the animal gained energy at a rate E, determined by the diameter of the foraging range (d) and the density of prey (a): E=

ei − ew tw (d, a) − ep tp (d, a) tw (d, a) + tp (d, a)

,

(9.6)

where ei equals the energetic content of prey, ew equals the energetic cost of waiting, ep equals the energetic cost of pursuing, tw equals the time spent waiting, and tp equals the time spent pursuing. Buckley assumed animals maximize their rate of energy gain by foraging at an intermediate diameter (d∗ ). As the density of foragers (N) increases, the potential diameter of foraging becomes smaller than the energetically optimal diameter (d < d∗ ). This mechanism generates density-dependent growth of the population: N = [m · tf · E(d) − µ]N,

(9.7)

where µ equals the rate of mortality, tf equals the time spent foraging, and m converts energy into offspring. Using these relationships, Buckley solved for the equilibrium density of the population. Buckley used her model to predict current and future distributions of two species of lizards (Sceloporus graciosus and Sceloporus undulatus). Sceloporus graciosus occurs in western regions of the USA, whereas S. undulatus occurs throughout southeastern and central regions of this country. The extensive knowledge of the ecology and physiology of these species enabled Buckley to parameterize her mechanistic model. For example, energetic costs of waiting (ew ) and pursuing (ep ) were estimated as factors of the resting metabolic rate of each species. Prey densities were based on samples of insects from relevant locations. Operative environmental temperatures were calculated for the current climate and a hypothetical climate that warmed uniformly by 3◦ C. For most parameters, Buckley used mean values for each species or estimated values from data for a closely related species. For a few parameters of S. undulatus, she used a set of values representing lizards from different regions. By comparing the predictions derived from this

222

T H E R M A L A D A P TAT I O N

(a)

Collection record Model prediction 95% CI (b)

Current range s6w2 s2w6 Figure 9.3 A mechanistic model was used to predict the northern limits of the sachem skipper (Atalopedes campestris) in current and future climates. (a) The predicted northern limit of the current range in relation to collection records. Dotted lines show the 95% conﬁdence interval (CI) around the predicted range, derived from bootstrapping empirical data to generate different parameter values. (b) The predicted northern limit after 50 years for each of two warming scenarios: (1) 6◦ C per century in summer and 2◦ C per century in winter (s6w2), and (2) 2◦ C per century in summer and 6◦ C per century in winter (s2w6). For these predictions, the modelers assumed that (1) the winter survivorship was a logistic function of temperature and (2) the net reproductive rate depended on the thermal sensitivities of survivorship, fecundity, and voltinism. Reproduced from Crozier and Dwyer (2006, The American Naturalist, University of Chicago Press). © 2006 by The University of Chicago.

set, we can see how geographic variation in phenotypes might inﬂuence the response to climate change. For example, phenotypes in New Jersey were predicted to shift farther during warming than were phenotypes in Nebraska, largely because of a difference in body size (Fig. 9.4). Nevertheless, both parameterizations of the mechanistic model

predicted more extensive shifts in the northern limit than did a correlative model based on thermal variables (Fig. 9.4). Because the mechanistic model generally predicts the current range more accurately than the correlative model does, we should heed the discrepancy between their predictions about the future.

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

223

(a) Mechanistic model Current climate

Warming scenario

NE

NJ

(b) Correlative model Current climate

Lizards/1000 m2 <150

Warming scenario

300

450

600

Figure 9.4 (a) A mechanistic model was used to predict the geographic distributions of the eastern fence lizard (Sceloporus undulatus) in current and future climates. A uniform warming of 3◦ C was applied to current temperatures to approximate a future climate. The predictions shown here were based on phenotypes in Nebraska (NE) and New Jersey (NJ). (b) For comparison, geographic distributions were also predicted by a correlative model (Genetic Algorithm for Rule-set Production). Adapted from Buckley (2008, The American Naturalist, University of Chicago Press). © 2008 by The University of Chicago.

9.3.3 Predicting differential responses of populations and species Mechanistic models not only help us to infer distributional impacts of climate change, they can also reveal how adaptation to current environments determines the response to future environments. In

the models described above, all individuals of a species shared the same thermal physiology; this assumption seems naïve given the focus of this book. Contrary to what the poet Gertrude Stein once penned, a rose is not a rose is not a rose. Rather, all species exhibit some genetic variation

T H E R M A L A D A P TAT I O N

in thermal physiology, which should cause populations to respond differently over space and time, even if their environments warm similarly. In the example provided by Buckley (2008), we saw how geographic variation in body size affected the predicted response to climate change. Let us extend this focus to include geographic variation in thermal sensitivity. To illustrate the problem, imagine we wish to parameterize a mechanistic model for a species that occurs throughout a broad latitudinal range. We could sample populations throughout this range, and use the mean values of each trait. However, we would probably ﬁnd that certain parameters vary with latitude. Based on simple optimality models, the thermal optimum of performance should decrease as we sample populations from higher latitudes (see Fig. 3.19). We might also observe a corresponding increase in performance breadth. Which values of these parameters should we use to predict the response to climate change? Actually, choosing any one value would be a bad idea because local responses depend on this variation in thermal physiology. Deutsch and colleagues (2008) illustrated the signiﬁcance of heeding variation in thermal physiology when predicting the impacts of climate change. Using published data, they estimated the thermal sensitivities of ﬁtness (r) in 46 species of insects. These ﬁtness curves were used to calculate two indices: the warming tolerance and the safety margin. The warming tolerance was deﬁned as the difference between the critical thermal maximum and the mean environmental temperature. The safety margin was deﬁned as the difference between the thermal optimum for performance and the mean environmental temperature. Because both indices depend strongly on the seasonality of the environment, they both increase with increasing latitude. A similar relationship between seasonality and warming tolerance was found to exist among species of lizards, frogs, and turtles. Consequently, identical warming of tropical and temperate environments will have different biological consequences. Even when these researchers considered spatial heterogeneity of warming (scenario A2; IPCC 2007), tropical insects were predicted to suffer a greater reduction in ﬁtness than were temperate insects

0.3 0.2 Change in fitness (r)

224

0.1 0.0 –0.1 –0.2 –0.3 60S

40S

20S

0

20N

40N

60N

Latitude Figure 9.5 Environmental warming during the twenty-ﬁrst century was predicted to increase the ﬁtness (r ) of temperate insects but decrease the ﬁtness of tropical insects. Change in ﬁtness was calculated using temperatures observed for 1950–1990 and those predicted for 2070–2100. Fitness at each time was calculated by integrating ﬁtness curves over diel and seasonal cycles of temperature. The solid line shows a spline ﬁt to the data. Adapted from Deutsch et al. (2008) with permission from The National Academy of Sciences. © 2008 by The National Academy of Sciences, USA.

(Fig. 9.5). Relaxing assumptions about seasonal activity and thermoregulatory behavior did not alter this prediction. Despite the fact that northern environments will continue to warm more rapidly than the tropics, tropical species may face the greatest risk of extinction. In the same vein, we should expect the most severe phenotypic selection to occur in tropical environments. Thus, the actual risk of extinction will depend on the demographic beneﬁt of adaptation versus the demographic cost of environmental change. Given this reality, let us now consider the possibility of adaptation to global warming.

9.4 Adaptation to directional thermal change Up to this point, we have not considered how adaptation might mitigate the impacts of global warming. Although the primary response during glacial retreats seems to have been shifts in geographic ranges (Parmesan 2006; Parmesan and Galbraith 2004), current barriers to dispersal may

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

selection alters thermal physiology. Such physiological adaptations are more difﬁcult to detect than phenological adaptations because historical records of thermal physiology pale in comparison to records of phenological events. Still, indirect evidence exists for rapid evolution of thermal physiology. Urbanization has warmed local environments as fast as or faster than greenhouse gases have warmed our planet (Oke 1997). Not surprisingly, the performance curves of several species of fungi differed between urban and rural genotypes (Fig. 9.6). Genetic markers associated with the heat tolerance of Drosophila subobscura have tracked climate warming on multiple continents (Balanya et al. 2006). Experimental studies of physiological adaptation have complemented these comparative studies. After translocating plants (Umbilicus rupestris) north of their range limit, Woodward

force adaptation or extinction. Given the range of adaptations discussed in this book, we should question efforts to predict biological responses in the absence of evolution. Indeed, based on the wealth of evidence for thermal adaptation within and among species, we should expect adaptation to current warming (Holt 1990; Jump and Penuelas 2005; Parmesan 2006; Skelly et al. 2007). Two kinds of adaptations to recent warming have been documented. First, natural selection alters phenological mechanisms, such as photoperiodic thresholds for activity and dormancy (reviewed by Bradshaw and Holzapfel 2006). For example, the photoperiodic threshold for diapause has evolved in the fall webworm (Hyphantria cunea) within the last 15 years (Gomi et al. 2007); worms collected in 2002 entered diapause at a shorter photoperiod than did worms collected in 1988 and 1995. Second, natural

Diameter of growth (mm)

20

16 Chrysosporium pannorum

Torulomyces lagena 14

18

12

16

10 14

8

12 10

6 16

18

20

22

24

26

28

4

16

Trichoderma koningii

20

22

24

26

28

22

24

26

28

Penicillium bilaii

40

28

30

24

20

20

10

16

0

18

32

50 Diameter of growth (mm)

225

16

18

20

22

24

Temperature (°C)

26

28

12

16

18

20

Temperature (°C)

Figure 9.6 Thermal sensitivities of growth rate differed between urban and rural isolates of chitinolytic fungi. Urban isolates of Chrysosporium pannorum and Trichoderma koningii grew faster than rural isolates at 26◦ C, but grew slower than rural isolates at 18◦ C. In Penicillium bilaii and Torulomyces lagena, urban isolates grew as fast as or faster than rural isolates at each temperature. Urban and rural isolates are denoted by open and ﬁlled symbols, respectively. Error bars depict 95% conﬁdence intervals. Adapted from McLean et al. (2005) with permission from Elsevier. © 2005 Elsevier Ltd.

226

T H E R M A L A D A P TAT I O N

(1990) observed the evolution of greater cold tolerance in just 9 years. Experimental warming of aquatic mesocosms also caused the rapid evolution of thermal physiology (Box 9.1). Clearly, we can expect both phenological and physiological adaptations to result from anthropogenic changes in temperature. As we have discussed throughout this book, the rate of adaptation depends on many factors, including phenotypic plasticity, genetic variation, and evolutionary tradeoffs. Adaptive phenotypic plasticity, such as thermoregulatory behavior or physiological acclimation, will limit the strength of phenotypic selection (Huey et al. 2003; Jump and Penuelas 2005). However, these forms of plasticity impose costs (see

Chapters 4 and 5). Furthermore, acclimation capacities should be most limited in tropical organisms (see Stillman 2003), which might face the greatest threat during climate change (Deutsch et al. 2008). The evolutionary response to phenotypic selection requires genetic variation. Such variation will deteriorate under strong selection, slowing the rate of adaptation (Bell and Collins 2008). Moreover, the death of individuals caused by selection will enhance genetic drift, leading to further loss of genetic variance (Holt 1990; Parmesan 2006). Immigration could enhance or inhibit the response to selection, depending on the ﬁtness of the incoming genotypes (see Chapter 3). Finally, tradeoffs between traits can constrain the course of thermal adaptation; we can characterize

Box 9.1 Experimental warming in mesocosms Although we generally think of adaptation as a process that requires considerable time, some organisms will adapt rapidly to changes in temperature. Van Doorslaer and colleagues (2007) demonstrated this fact by experimentally warming aquatic communities. They established 24 mesocosms containing sediment, plankton, ﬁsh, and nutrients. These mesocosms were exposed to either ambient

temperatures or elevated temperatures. The two elevated thermal treatments corresponded to a likely scenario of warming during the next century (A2; IPCC 2007) and an additional increase of 50% (A2 + 50%). After 1 year, a species of water ﬂeas (Simocephalus vetulus) was sampled from each of the mesocosms for a study of thermal reaction norms (Fig. 9.7). By raising two clones from each mesocosm

Three thermal treatments

Three incubation temperatures Clonal sublines

Ambient A2

A2 + 50%

18°C

Repeat 1

22°C

Repeat 2

26°C

Repeat 3

Two clonal lines

Replica 1 Replica 2

20°C

Replica 3 Two generations Replica 4

Figure 9.7 The design of a warming experiment conducted with aquatic mesocosms. From left to right, the diagram depicts the initial exposure of communities to several selective regimes and the subsequent sampling of water ﬂeas (Simocephalus vetulus) for laboratory studies of thermal reaction norms. Reproduced from Van Doorslaer et al. (2007) with permission from Blackwell Publishing. continues

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

227

Box 9.1 Continued at three temperatures, the researchers documented genetic changes in thermal sensitivities. The most striking ﬁnding was an evolutionary increase in the thermal tolerance of S. vetulus in heated environments. Speciﬁcally, clones from the heated mesocosms survived better at 26◦ C than did clones from the unheated mesocosms (Fig. 9.8). No apparent tradeoffs with survivorship at low temperature occurred.

Still, clones from all three mesocosms attained similar rates of population growth (r ) at all three temperatures, suggesting some tradeoff was associated with the evolution of thermal tolerance. Although heat tolerance clearly evolved during warming, we must wonder whether adaptation would enable this population to persist given that the change in survivorship failed to affect ﬁtness.

Ambient mesocosm temperature Mesocosm temperature A2 Mesocosm temperature A2 + 50%

Survivorship (%)

100 80 60 40 20 0

18 22 26 Incubation temperature (°C)

these tradeoffs through studies of genetic covariances (see Chapter 1). The failure to address thermal adaptation limits the generality of current models, whether they are correlative or mechanistic. Given the evidence on hand, it seems silly to ask whether species will adapt to anthropogenic warming. Rather, we should ask different questions. How quickly and in what ways will certain species adapt? Which species will adapt too slowly to avoid extinction? To answer these questions, we need to consider slightly different models than those presented in other parts of this book. In most models of thermal adaptation, the environmental temperature changes cyclically or stochastically. We need models in which the environmental temperature changes directionally. For convenience, I shall cast these models in terms of environmental warming, but the models apply equally well to environmental cooling.

Figure 9.8 Experimental warming caused the evolution of thermal reaction norms for survivorship in Simocephalus vetulus. At 26◦ C, clones from heated mesocosms were more likely to survive from birth until the date of producing a second clutch of offspring. Error bars depict standard errors of the means. Adapted from Van Doorslaer et al. (2007) with permission from Blackwell Publishing.

9.4.1 Adaptation of thermoregulation Many organisms rely primarily on thermoregulation to deal with thermal change (see Chapter 4). The ﬂexibility of behavioral and physiological mechanisms of thermoregulation enables organisms to respond rapidly to warming. In tropical environments, both ectotherms and endotherms can ameliorate the negative physiological consequences of warming (Hertz 1981). At higher latitudes, ectotherms could thermoregulate to increase physiological performance during a warmer spring or avoid energetic stress during a milder winter. For endotherms, the warming at high latitudes should inﬂuence strategies of endothermy and torpor (Humphries et al. 2002). As in the current climate, thermoregulation should confer obvious beneﬁts in future climates. But how will warming affect the costs of thermoregulation?

228

T H E R M A L A D A P TAT I O N

During global warming, opportunities to achieve preferred body temperatures will shift both dielly and seasonally. Quantifying these shifts requires detailed models of the physical environment (see Chapter 2). Recently, Michael Sears (unpublished observations) modeled the availability of preferred microclimates for a small lizard in the Mescalero Sand Dunes (New Mexico, USA) during a typical day in June. To simulate the effect of global warming, he then added 3◦ C to all operative temperatures and recalculated the available habitat. In the contemporary climate, habitat was most accessible in the early morning; only a few small patches remained usable by early afternoon (Fig. 9.9a). Under the global warming scenario, the abundance of preferred microclimates peaked about 40 minutes earlier in the day. Usable habitat became more limited and more fragmented at 0900 h, and disappeared completely by 1300 h (Fig. 9.9b). Because operative and preferred temperatures vary among species, temporal shifts in available habitat will differ for predators and prey. Any degree of warming can alter the potential overlap between predators and prey, as was predicted for ﬁsh and invertebrates in several lakes (Jansen and Hesslein 2004).

These modeling exercises tell us about the physical constraints on thermoregulation, but we still need to infer the costs of potential behaviors. With these costs, we can link temporal changes in thermal microclimates to adaptive shifts in thermoregulatory strategies. Remember, the cost of behavioral thermoregulation depends on the spatial distribution of operative temperatures, as well as their statistical distribution (see Box 4.4). Any scenario of warming changes not only the frequencies of temperatures but also the structure of the environment from the animal’s perspective. Consider the example provided by Sears (Fig. 9.9). Applying a reasonable warming increased the distance between preferred microclimates in the early morning. This phenomenon would likely increase the energetic cost of behavioral thermoregulation. The exact increment in cost depends on the properties of the organism, such as mobility, perception, and memory. If the new cost of thermoregulation outweighs the beneﬁt, we should expect the organism to thermoregulate poorly during activity or to cease activity altogether. Importantly, modest warming on a global scale could cause complex changes in behavior on local scales. These changes

(a) Current climate 0900

1300

(b) Warming scenario (+3°C) 0900

1300

Figure 9.9 Global warming will alter the spatial structure of thermal environments. These maps depict the availability of preferred microclimates of Sceloporus arenicolus (shaded pixels) in the Mescalero sand dunes of New Mexico, USA (M. W. Sears, unpublished observations). Preferred microclimates include operative temperatures in the range of 29–37◦ C. In the current climate, the distribution of preferred microclimates becomes extremely rare by 1300 h. Awarming of 3◦ C would reduce preferred microclimates at 0900 h and eliminate them completely at 1300 h, potentially shifting activity to an earlier time of day.

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

will surely exert selective pressures on the thermal sensitivities of physiological performances.

9.4.2 Adaptation of the thermal optimum How would gradual warming inﬂuence the evolution of performance curves? We can address this question with a model developed by Lynch and Lande (1993), who studied adaptation to directional environmental change. This model strongly resembles the one developed by Lynch and Gabriel (1987), which was covered in Chapter 3; however, Lynch and Lande directly modeled the environmental sensitivity of ﬁtness. For a growing population, the instantaneous rate of increase of a genotype (r) provides an appropriate estimate of ﬁtness (see Box 3.3). Lynch and Lande modeled ﬁtness using a symmetrical function, which we shall interpret as a thermal reaction norm: (z − Te )2 r(z, Te ) = rm − , (9.8) 2σ 2 where rm equals the maximal intrinsic rate of increase, z equals the thermal optimum, σ equals the thermal breadth, and Te equals the operative environmental temperature. Clearly, this equation describes a perfect thermoconformer; to apply the model to a thermoregulator, one would simply replace the operative temperature (Te ) with the body temperature (Tb ). The mean and variance of the thermal optimum (z and Vz ) depend on the additive genetic effect ( g, with variance Vg ), as well as environmental and nonadditive effects (e). Assuming e equals zero and has a variance of Ve , the growth rate of the population equals (¯g − Te )2 + Vz , (9.9) 2σ 2 where Vz = Vg + Ve . As in many models of evolution, a constant additive genetic variance was assumed. Note that the mean phenotypic value (¯z) equals the mean genotypic value (¯g). Lynch and Lande assumed the environment changes at a constant rate (k), with some stochastic variation (ε) around this trend: r(z, Te ) = rm −

Te (t) = kt + ε.

(9.10)

The stochastic component has a mean of zero and a variance equal to VT . During environmental warming, the ﬁtness of a previously adapted genotype

229

declines and selection drives the thermal optimum toward the new operative temperature. Because the environment continues to warm, the thermal optimum will always lag behind the mean operative temperature. After some time, a balance between environmental change and natural selection enables the population to persist. If the environment warms too rapidly, however, this balance never occurs and the population declines to extinction. To what extent can adaptation of the thermal optimum enable a population to persist during environmental warming? To answer this question, Lynch and Lande calculated the expected value of r for the population at an equilibrium maintained by selection, drift, and mutation. Obviously, this rate of growth must exceed zero for the population to persist. Ideally, we should like to know the exact outcome of evolution in a warming environment. Unfortunately, environmental stochasticity and genetic drift prevent us from knowing this outcome in advance. Therefore, the predictions of interest become the probability distributions of the thermal optimum (z) and the rate of increase (r) at the evolutionary equilibrium. We can calculate the expected thermal optimum, E(¯g), and its variance, V(¯g), from the following equations: E(¯g) = kt − k

σ2 Vg

(9.11)

and V(¯g) =

Vg VT σ2 + . 2Ne 2σ 2

(9.12)

Equation 9.11 tells us that either a large performance breadth or a low genetic variance would cause the thermal optimum to lag far behind the operative temperature. Equation 9.12 tells us the outcome of evolution would be less predictable in smaller populations and more variable environments. Most of these insights should accord with our intuition, but let us brieﬂy focus on one of them: the inﬂuence of the performance breadth on the expected thermal optimum. All other things being equal, a warming environment affects a thermal specialist more than it affects a thermal generalist. Thus, the thermal optimum of a generalist undergoes weaker selective pressure and lags farther behind the operative temperature.

230

T H E R M A L A D A P TAT I O N

Once we know the expected thermal optimum, we can calculate the expected growth rate of the population, E(r), at the evolutionary equilibrium: Vz k2 σ 2 1 VT Vg E(r) = rm − − − − + 1 . 2σ 2 2(Vg )2 4Ne 2σ 2 2σ 2 (9.13) From this equation, we see that the expected growth rate can never reach the maximal growth rate (rm ) because of directional and stochastic changes in temperature (k and VT , respectively). The degree of maladaptation to directional change depends on the performance breadth (σ ) and genetic drift (∝ 1/Ne ). The degree of maladaptation to stochastic change depends on the performance breadth and the additive genetic variance (Vg ). Equation 9.13 enables us to calculate the maximal rate of warming that permits the persistence of the population (Lynch and Lande 1993). Not surprisingly, this critical rate (kc ) depends on the same key factors that deﬁne the expected thermal optimum: 1/2 Vg Vg 1 V kc = 2¯rm − − T2 + 1 , σ 2Ne σ 2σ 2 (9.14) where r¯m equals the rate of population increase when the mean thermal optimum matches the mean operative temperature. Therefore, the likelihood of persistence under directional thermal change increases as (1) the genetic variance of the thermal optimum increases, (2) the performance breadth decreases, (3) the population size increases, and (4) the stochasticity of temperature decreases. In a ﬁnite population, the risk of extinction during constant warming always exceeds that predicted by a deterministic model. For this reason, Burger and Lynch (1995) extended the model of Lynch and Lande (1993) to explore the effects of stochastic processes on the risk of extinction. This exercise required computer simulations under a set of assumptions that corresponded closely to those of Lynch and Lande’s model. Burger and Lynch considered a ﬁnite population of asexual, semelparous organisms with nonoverlapping generations. From our perspective, the phenotype under selection was the thermal optimum for survivorship; all individuals produced the same number of offspring if they survived to maturity. If the number of surviving offspring exceeded a

hypothetical carrying capacity, excess individuals were culled randomly before mating. The thermal optimum was determined by additive genetic and environmental effects, whose variances possessed the same properties as they did in Lynch and Lande’s deterministic model. By coupling environmental and demographic stochasticities, Burger and Lynch quantiﬁed the way that adaptation affected the probability of persistence during directional environmental change. The distinction between this exercise and the one undertaken by Lynch and Lande (1993) should not be overlooked. Because Lynch and Lande derived the critical rate of environmental change for the expected phenotype, E(r), their predictions hold only in a deterministic world. Speciﬁcally, their model predicts a population would persist unless the rate of environmental change exceeds the critical rate (kc ). But stochasticity dictates that an environment warming slowly over a long period can warm rapidly over a short period. A chance stretch of very warm years could lead to extinction even when the mean rate of change falls short of kc . Furthermore, demographic stochasticity caused by density dependence can lead to maladaptation via genetic drift. The unpredictability of the real world makes the persistence of a population in a warming environment far more questionable. In this case, extinction always occurs if we consider a sufﬁcient period of time. But we can ask the following question: what is the probability of extinction during a certain period of warming? Computer simulations conﬁrmed that environmental and demographic stochasticities drive a population to extinction even when the environment warms slower than kc (Fig. 9.10). In each simulation, the mean phenotype eventually lagged behind the optimal phenotype and the population shrank. As each population dwindled through natural selection, genetic drift began to erode additive genetic variance, which in turn caused future maladaptation and more severe selection. Depending on the rate of environmental change, each population would either rebound or become extinct. The take-home message is clear. A real population, which experiences selection and drift in a stochastic environment, faces a greater risk of extinction than does an idealized population in a deterministic environment.

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

(a) 80 60

Number of simulations

40 20 0 0

60

120

180

240

0

3000 6000 9000 12 000 Generations before extinction

300

(b) 60 50 40 30 20 10 0 15 000

Figure 9.10 In simulations of evolution during directional environmental change, the likely time until extinction depended on the rate of change. (a) During rapid change (k = 0.12), the majority of populations were extinct within 100 generations. (b) During slow change (k = 0.06), populations persisted for much longer periods. Adapted from Burger and Lynch (1995) with permission from Blackwell Publishing.

9.4.3 Adaptation of the performance breadth Huey and Kingsolver (1993) extended Lynch and Lande’s analysis to consider two additional questions. First, how does directional change in temperature affect the optimal performance breadth? And second, how do specialist–generalist tradeoffs inﬂuence adaptation to directional change? To answer these questions, they permitted the performance breadth (σ ) to evolve in addition to the maximal rate of increase (rm ). In one scenario, the evolution of σ imposed no effect on the evolution of rm . In a second scenario, an increase in σ caused a decrease in rm . With these minor modiﬁcations to the deterministic model, Huey and Kingsolver calculated the performance breadth that maximized the critical rate of warming (kc ).

231

Interestingly, genotypes with intermediate performance breadths would tolerate the fastest change in temperature (Huey and Kingsolver 1993). This result makes sense if we consider the details of the model. From eqn 9.11, we concluded that the thermal optimum of a generalist would lag farther behind the operative temperature than the thermal optimum of a specialist. Hence, a population of generalists has difﬁculty persisting during rapid warming (see eqn 9.14). Nevertheless, specialists face a different problem. If their performance breadth is too narrow, even a slow rise in temperature would lead to extinction. Consequently, an intermediate performance breath balances the needs to conserve ﬁtness and facilitate adaptation during environmental change. A specialist–generalist tradeoff exacerbates the problem of a wide performance breadth because performance at the thermal optimum decreases as the performance breadth increases. Adding environmental or demographic stochasticity to the model does not alter the basic result: genotypes with an intermediate performance breadth fare best in a changing environment. Huey and Kingsolver (1993) added stochasticity to the directional change in temperature. As thermal stochasticity increased, so did the performance breadth needed to maximize the critical rate of warming. Similarly, Burger and Lynch (1995) modeled the probability of extinction under environmental and demographic stochasticity. An intermediate performance breadth generally maximized the expected time until extinction (Fig. 9.11). The causes of these patterns differ slightly because of the genetic assumptions of each model. Huey and Kingsolver assumed a constant additive genetic variance, whereas Burger and Lynch modeled changes in genetic variance with a multilocus system of alleles. Therefore, the simulations of Burger and Lynch also revealed the potential for genetic constraints to limit adaptation. Under the strong selection resulting from a narrow performance breadth, a decrease in the additive genetic variance slowed the rate of adaptation and increased the probability of extinction. In fact, wider performance breadths always enhance persistence if they increase the additive genetic variance of the thermal optimum (Huey and Kingsolver 1993).

232

T H E R M A L A D A P TAT I O N

predicted by models of adaptation to stochastic change.

105

Mean time to extinction

Slow change 104

9.5 Thermal games in a warming world

103

102

Rapid change 101

1

2

3

4 5 6 7 8 Breadth of fitness (s)

9

10

Figure 9.11 For most rates of warming, an intermediate thermal breadth would maximize the persistence of a population. Mean times to extinction were derived from simulated adaptation to various rates of directional environmental change (k = 0.055 − 0.250). If an environment warms rapidly, only an extremely wide thermal breadth would enable a population to persist for more than a few generations. Adapted from Burger and Lynch (1995) with permission from Blackwell Publishing.

All things considered, both the thermal optimum and the performance breadth inﬂuence the viability of a population during environmental warming. This conclusion bolsters an earlier one based on a model described in Chapter 3. Speciﬁcally, Gilchrist (1995) found that generalists were more ﬁt than specialists when temperature varied little within generations and greatly among generations. The seasonal variation among generations of Gilchrist’s model functionally equals the steady increase in temperature of Huey and Kingsolver’s model. The only difference being that the temperature rises and falls cyclically instead of rising continuously. Both patterns of thermal change favor generalists. Given this robust prediction, regions that have heated more rapidly in recent decades should contain genotypes that have wider performance breadths. In fact, global warming might explain why the performance breadths of real organisms often exceed those

In Chapter 8, we saw how biotic interactions can inﬂuence the course of thermal adaptation. What happens if we apply the concept of a thermal game to evolution in a warming environment? Consider a simple game between two or more competitors that was recently analyzed by Johansson (2008). With a few assumptions, this game can be used to infer the outcome of coevolution during warming. First, let us assume that the thermal sensitivity of ﬁtness16 follows a Gaussian function, and the environment conforms to a Gaussian distribution of operative temperatures. For simplicity, let us also assume that the competing species resemble one another physically, such that they experience identical operative temperatures in the same microhabitat. The presence of one species in a microhabitat prevents the other species from effectively using this site, thus altering the frequency of available temperatures (see Fig. 8.9). Thus, competition between species increases as their thermal optima become more similar; if they possess the same thermal optimum, the model effectively becomes an analysis of a single species. Given these assumptions, the evolutionarily stable strategy in a constant environment resembles the one described in Chapter 8 (see Fig. 8.10). As long as the range of operative temperatures exceeds the thermal breadth of ﬁtness, the two species can coexist. At an evolutionary equilibrium, their thermal optima will deviate symmetrically from the mean operative temperature. Species with lower or higher thermal optima would not perform well enough to invade the community, and those with a thermal optimum equal to the mean would face too much competition. As the thermal breadth of each species decreases, the number of coexisting species in the 16

More precisely, Johansson (2008) modeled the thermal sensitivity of the carrying capacity of a species, but I consider the carrying capacity to be an estimate of ﬁtness. This difference in terminology serves to simplify the explanation of the model without loss of generality.

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

evolutionarily stable strategy increases. For a community of three species, the thermal optimum of the third species would equal the mean operative temperature, while the thermal optima of the other two would deviate symmetrically from this temperature (as in the evolutionarily stable strategy for two species). How might species coevolve during global warming? To answer this question, we can draw on Johansson’s simulations of adaptation during directional environmental change. Johansson examined the coevolution of competing species using an individual-based model, which I interpret here in the language of thermal biology. First, Johansson let genetic variation build up during 20 000 generations of evolution in a constant environment. Then, he initiated environmental change by increasing the mean operative temperature at a constant rate. During warming, the thermal optima of the species were permitted to coevolve, but the thermal breadths remained constant. In simulations of a single species, the outcomes matched the prediction of Lynch and Lande’s model (see Section 9.4.2); speciﬁcally, the thermal optimum lagged behind the mean operative temperature, but the species persisted as long as warming proceeded slowly. In simulations of two competing species, one species was very susceptible to extinction during warming (Fig. 9.12). This result stemmed from the different phenotypes of the species at the onset of warming. Initially, one species had a thermal optimum that exceeded the operative temperature, while the other species had a thermal optimum that did not. Therefore, warming immediately increased the ﬁtness of one species while decreasing the ﬁtness of the other species (Fig. 9.12a). For the advantaged species, the thermal optimum evolved ahead of but close to the mean operative temperature. For the disadvantaged species, the thermal optimum lagged far behind. The fate of the lagging species depended on the rate of warming, but extinction was much more likely than it was without competition. In fact, the lagging species became extinct at a rate of warming that was only one-ﬁfth of the critical rate for a single species! When the community initially consisted of three species, the species with the lowest thermal optimum became extinct unless the environment warmed extremely slowly (< 10% of the critical

233

rate for a single species; Fig. 9.12c and d). Upon extinction of the third species, the remaining two evolved thermal optima that tracked the mean operative temperature (Fig. 9.12d). Taken together, these simulations tell us that competition will exacerbate risks of extinction during global warming. Intuition suggests predation or parasitism would do the same.

9.6 Evolutionary consequences of gene ﬂow in a warming world Adaptation to constant warming depends not only on natural selection at the local scale but also on gene ﬂow at the regional scale. Most species constitute a metapopulation, within which individuals migrate between locations with different thermal conditions. In Chapter 3, we brieﬂy considered the role of gene ﬂow during the evolution of performance curves (Section 3.9). Gene ﬂow initially enhances adaptation to a novel environment, but ultimately prevents perfect adaptation. However, these models focus on the migration of genotypes from a source population to a sink population, as occurs during a range expansion. During climate change, local genotypes become maladapted, but preadapted genotypes might exist in some part of the geographic range (see, for example, Rehfeldt et al. 2001). The balance between the immigration of preadapted and maladapted genotypes determines the impact of gene ﬂow on the adaptation and persistence of a population. In this ﬁnal section, we shall examine the interaction between natural selection and gene ﬂow in the context of global warming.

9.6.1 Spatially heterogeneous warming can reduce the ﬂow of maladapted genotypes A model developed by Kirkpatrick and Barton (1997) suggests that global warming will alter the effect of gene ﬂow on local adaptation. Instead of focusing on environmental variation over time, Kirkpatrick and Barton focused on environmental variation over space. Let us assume that the operative environmental temperature varies linearly along a spatial gradient (e.g., latitude), such that the temperature at any location (x) can be

234

T H E R M A L A D A P TAT I O N

(a) 3.0

(b)

Thermal optimum

2.5 2.0 Species 1

1.5 Species 1

1.0 0.5 0 –0.5

Species 2

Species 2

–1.0 (c) 3.0

(d)

2.5 Species 1

Thermal optimum

2.0

Species 1

1.5 1.0

Species 2

0.5

Species 2

0 Species 3

–0.5 –1.0 0

Species 3 0.2 0.4

0.6 0.8 1 1.2 1.4 1.6 1.8 Generations (× 105)

2

0

0.2 0.4

0.6 0.8 1 1.2 1.4 1.6 1.8 Generations (× 105)

2

Figure 9.12 Competition inﬂuenced the simulated adaptation of a continuous trait, equivalent to the thermal optimum, during directional environmental change. (a) In a community of two species, the thermal optimum of one species (lower, solid line) lags behind the mean temperature (middle, dashed line), while the thermal optimum of the other species (upper, solid line) adapts to the changing environment. (b) If warming occurs more rapidly, the maladapted species may become extinct. The different responses of the two species occur because the ﬁtness of one species initially increases during warming while that of the other species initially decreases. (c and d) In a community of three species, either the thermal optima of all three species parallel the mean temperature (c) or one species becomes extinct relatively soon after warming commences (d). Adapted from Johansson (2008) with permission from Blackwell Publishing.

calculated as Te (x) = ax + b,

(9.15)

where a and b represent constants. Ametapopulation distributed along this cline would undergo thermal adaptation while exchanging genes by migration. To model local adaptation, Kirkpatrick and Barton used the same ﬁtness function as did Lynch and Lande (eqn 9.8). Gene ﬂow occurred via random dispersal prior to reproduction. A simple model of population dynamics linked adaptation to the density of individuals at each location, N(x): N(x) = cer¯ d ,

(9.16)

where c and d represent constants. When d = 0, eqn 9.16 imposes a form of density dependence in which the population consists of c individuals at its equilibrium. Otherwise, the density at the equilibrium depends on the population’s mean ﬁtness (¯r). Depending on the magnitude of the thermal cline, the model predicts one of two outcomes (Fig. 9.13). A shallow cline enables a species to expand its range along the entire gradient and adapt perfectly to each local environment. In contrast, a steep cline limits the geographic range and prevents perfect

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

Moderate cline

Shallow cline

Density

Steep cline

235

z or Te

10

Te

0

Te Te, z

z

z

–10 –5

0

5

–5

0 Location (x)

5

–5

0

5

Figure 9.13 The steepness of a thermal cline greatly affects the predicted geographic distribution of a species. Bottom plots show the predicted clines in the thermal optimum, z (solid lines), for different clines in environmental temperature, Te (dashed lines). Upper plots show the corresponding distributions of individuals. Along a shallow thermal cline, a species readily adapts to local environments and occupies the entire cline. As the thermal cline becomes steeper, the inability to adapt to local conditions constrains the range of the species. Adapted from Kirkpatrick and Barton (1997, The American Naturalist, University of Chicago Press). © 1997 by The University of Chicago.

adaptation at each location. The density of individuals follows a Gaussian distribution, with the greatest density occurring at the center of the range (Fig. 9.13). Gene ﬂow from the center of the range limits adaptation at the edge. Kirkpatrick and Barton referred to the range boundary as soft because some number of individuals technically exists at any point along the linear gradient; yet reproductive constraints and demographic stochasticity would generate a hard range boundary. In a more complex model of population dynamics, gene ﬂow can also cause extinction when the thermal cline becomes too steep relative to the thermal breadth of the species (Kirkpatrick and Barton 1997). Bear in mind that this model assumes the thermal breadth cannot evolve. How might global warming affect the interaction between natural selection and gene ﬂow? Kirkpatrick and Barton’s model offers a tentative answer to this question. When gene ﬂow limits adaptation, a change in the thermal cline would perturb the metapopulation from its equilibrial state. Imagine a homogeneous warming along the linear gradient (i.e., an increase in b of eqn 9.15). The range of the species would shift along the gradient, but the distribution of densities and the degree of adaptation should remain the same (Kirkpatrick and Barton 1997). But now imagine that warming occurs heterogeneously along the linear gradient. This phenomenon would alter the steepness of the thermal cline (i.e., change a as well as b). In the

real world, latitudinal clines in temperature should become less steep because environments at high latitudes continue to warm faster than those at low latitudes (see Section 9.1). This dampening of thermal clines could facilitate adaptation to local environments. Of course, many other abiotic and biotic factors will change also, such that we cannot know whether the mechanism described by this model will actually enable species to persist (for a review of biotic limits to geographic ranges, see Case et al. 2005).

9.6.2 Spatially heterogeneous warming can increase the ﬂow of preadapted genotypes When environments warm at different rates, migration might enhance the rate of local adaptation. As the local environment warms, migration enables some individuals to escape local stress and ﬁnd an environment to which they are preadapted. This shufﬂing of genotypes can maintain genetic diversity and speed local adaptation. This process was recently modeled by de Mazancourt and colleagues (2008). They created environments consisting of discrete patches, each of which underwent environmental change at a different rate. Individuals of one or more species inhabited these patches, where they grew for some period and then reproduced according to their biomass. The offspring were assigned a distribution of phenotypes with a mean equal to

236

T H E R M A L A D A P TAT I O N

the mean parental phenotype in the patch. Offspring dispersed randomly among patches and the cycle continued. This process was simulated for communities of various sizes, ranging from one to 16 species. The growth rate (and hence the ﬁtness) of each individual depended on a Gaussian performance curve, similar to eqn 3.5. If we interpret this curve as the thermal sensitivity of growth rate, the model predicts the evolution of the thermal optimum in a warming environment. The thermal optimum of each species evolved by hard and soft selection. Hard selection occurred because the local environment determined the fecundity of each individual. Soft selection occurred because no patch could hold more than 10 000 individuals. If the number of residents plus immigrants exceeded this maximum, excess individuals were randomly culled from the population. Thus, recruitment of each phenotype in a patch was proportional to its abundance after dispersal. The performance breadth could not evolve, but a range of breadths were examined to see how this parameter inﬂuenced adaptation of the thermal optimum. At the beginning of each simulation, populations were allowed to adapt to a constant environment for 400 generations. Afterward, the researchers imposed 50 generations of spatially heterogeneous warming and 50 generations of recovery in a constant environment (warmer than the initial environment). In the absence of competition, heterogeneous warming enhanced persistence and sometimes sped adaptation. Under certain conditions, patches in which individuals were initially adapted provided stepping stones for adaptation to other patches. Under other conditions, preadaptation coupled with migration slowed adaptation to novel environments. Overall, the effect of spatial heterogeneity depended on the dispersal rate and the performance breadth. When the performance breadth was wide and the dispersal rate was low, increasing spatial heterogeneity increased the rate of adaptation. Otherwise, adaptation decreased with increasing spatial heterogeneity; very high dispersal led to maladaptation in all patches. In general, competition among species negated the beneﬁts of migration during environmental change, increasing the chance of extinction and

slowing the rate of adaptation. Individuals fared well in patches where they were preadapted, but were excluded from patches in which competitors were preadapted. The only exception occurred when performance breadths were narrow; in this case, extinction during warming was so severe that species rarely encountered one another. Both migration and competition combined to slow adaptation. High dispersal enabled species to sort themselves into patches where they were preadapted. In contrast, low dispersal forced species to adapt to changes in the local environment. Simulations of evolution in more complex communities could change our perception of the interaction between natural selection and gene ﬂow during global warming.

9.7 Conclusions Anthropogenic climate change poses a tremendous risk to biodiversity. Rates of warming during recent decades have exceeded those experienced during the previous millennia. These changes have led to advances in phenology, shifts in geographic ranges, and disruptions of ecological interactions. The current acceleration of global warming will exacerbate these effects in the coming decades. Mechanistic models provide a means to understand ecological impacts because they force researchers to make explicit assumptions about the biology of organisms. One can relax these assumptions to explore the relative contribution of different factors to changes in population dynamics and geographic distributions. Given this property, mechanistic models can ultimately be extended to deliver predictions about evolutionary impacts. A growing body of evidence underscores the need to consider evolutionary responses to global warming. Both the physiological regulation of phenology and the thermal sensitivity of performance have evolved in warming environments. Nevertheless, quantitative genetic and allelic models suggest that rapid warming will lead to persistent maladaptation and certain extinction. The degree of maladaptation during warming depends on numerous factors, including the stochasticity of temperature, the size of the population, the additive genetic variance, the rate of gene ﬂow, and the interactions between

A D A P TAT I O N TO A N T H R O P O G E N I C C L I M AT E C H A N G E

species. Although current models illustrate important constraints on adaptation, they are too simplistic to provide predictions about speciﬁc systems. By merging the ecological details of mechanistic models with the evolutionary processes of allelic models, we might arrive at an applied theory of thermal adaptation that more accurately predicts the impacts of anthropogenic warming. Three endeavors will determine the success of an applied theory. First, we must accurately characterize the selective environments. This task requires attention to spatial and temporal variations in the rate of thermal change. Both the mean and variance of temperature will change heterogeneously, and some environments in the future may not resemble any environment on earth today (Williams et al. 2007). Second, we must link dynamic changes in thermal environments to natural selection. A recurring theme of this book was the mismatch between theoretical predictions and empirical observations, even in cases where selective environments were known. Clearly, we need a better understanding of the tradeoffs that constrain phenotypes; otherwise, we will be stuck with poor representations of ﬁtness landscapes. Finally, we must accurately quantify the genetic constraints on thermal adaptation. In theory, directional selection, such as that caused by constant warming, should enhance the stability of

237

genetic variances and covariances ( Jones et al. 2004). In practice, we know far too little about the genetic architecture of thermal reaction norms to predict genetic constraints. Yet, these constraints ultimately determine whether adaptation will proceed rapidly enough to keep pace with environmental change. In the opening chapter, I articulated a challenge for evolutionary thermal biology: to explain the diversity of strategies by which organisms cope with their thermal environments. How will thermal biologists respond to this challenge? I hope we respond by establishing a strong connection between theoretical and empirical research. This book was designed to facilitate that connection. The models that we have examined constitute a foundational theory of thermal adaptation. This theory should guide us toward critical experiments, which would enable us to evaluate and reﬁne the models. The threat of rapid warming on a global scale reminds us of the challenges that remain and adds a sense of urgency to our mission. But we should not rush to simple conclusions or despair at slow progress. Given the thermal prognosis for this century, the need for an applied theory of thermal adaptation will certainly outlive our careers. Perhaps the next generation of researchers will achieve Levins’ vision by dissolving the boundary between theoretical and empirical biology.

References

Ackermann, M., and M. Doebeli. 2004. Evolution of niche width and adaptive diversiﬁcation. Evolution 58: 2599–2612. Addo-Bediako, A., S. L. Chown, and K. J. Gaston. 2000. Thermal tolerance, climatic variability and latitude. Proceedings of the Royal Society B 267:739–745. Adolph, S. C. 1990. Inﬂuence of behavioral thermoregulation on microhabitat use by two Sceloporus lizards. Ecology 71:315–327. Albentosa, M., R. Beiras, and A. P. Camacho. 1994. Determination of optimal thermal conditions for growth of clam (Venerupis pullastra) seed. Aquaculture 126:315–328. Almansa, E., J. J. Sanchez, S. Cozzi, C. Rodriguez, and M. Diaz. 2003. Temperature-activity relationship for the intestinal Na+ −K+ -ATPase of Sparus aurata. A role for the phospholipid microenvironment? Journal of Comparative Physiology B 173:231–237. Alonzo, S. H., and M. Mangel. 2001. Survival strategies and growth of krill: avoiding predators in space and time. Marine Ecology Progress Series 209:203–217. Alonzo, S. H., P. V. Switzer, and M. Mangel. 2003. Ecological games in space and time: the distribution and abundance of Antarctic krill and penguins. Ecology 84:1598–1607. Ames, D. 1980. Thermal environment affects production efﬁciency of livestock. Bioscience 30:457–460. Amo, L., P. Lopéz, and J. Martín. 2007a. Pregnant female lizards Iberolacerta cyreni adjust refuge use to decrease thermal costs for their body condition and cell-mediated immune response. Journal of Experimental Zoology 307A:106–112. —. 2007b. Refuge use: a conﬂict between avoiding predation and losing mass in lizards. Physiology & Behavior 90:334–343. Anderson, A. R., A. A. Hoffmann, and S. W. McKechnie. 2005. Response to selection for rapid chill-coma recovery in Drosophila melanogaster: physiology and life-history traits. Genetical Research 85:15–22. Anderson, D. R., K. P. Burnham, and W. L. Thompson. 2000. Null hypothesis testing: problems, prevalence, and an alternative. Journal of Wildlife Management 64:912–923.

238

Anderson, J. L., L. Albergotti, S. Proulx, C. Peden, R. B. Huey, and P. C. Phillips. 2007. Thermal preference of Caenorhabditis elegans: a null model and empirical tests. Journal of Experimental Biology 210:3107–3116. Andrews, R. M. 1998. Geographic variation in ﬁeld body temperature of Sceloporus lizards. Journal of Thermal Biology 23:329–334. Andrews, R. M., F. R. M. de la Cruz, and M. V. Santa Cruz. 1997. Body temperatures of female Sceloporus grammicus: thermal stress or impaired mobility? Copeia:108–115. Andrews, R. M., F. R. Mendez-de la Cruz, M. VillagranSanta Cruz, and F. Rodriguez-Romero. 1999. Field and selected body temperatures of the lizards Sceloporus aeneus and Sceloporus bicanthalis. Journal of Herpetology 33:93–100. Angilletta, M. J. 2001. Thermal and physiological constraints on energy assimilation in a widespread lizard (Sceloporus undulatus). Ecology 82:3044–3056. —. 2006. Estimating and comparing thermal performance curves. Journal of Thermal Biology 31:541–545. Angilletta, M. J., A. F. Bennett, H. Guderley, C. A. Navas, F. Seebacher, and R. S. Wilson. 2006a. Coadaptation: a unifying principle in evolutionary thermal biology. Physiological and Biochemical Zoology 79:282–294. Angilletta, M. J., and A. E. Dunham. 2003. The temperaturesize rule in ectotherms: simple evolutionary explanations may not be general. American Naturalist 162: 332–342. Angilletta, M. J., T. Hill, and M. A. Robson. 2002b. Is physiological performance optimized by thermoregulatory behavior?: a case study of the eastern fence lizard, Sceloporus undulatus. Journal of Thermal Biology 27:199–204. Angilletta, M. J., P. H. Niewiarowski, A. E. Dunham, A. D. Leache, and W. P. Porter. 2004b. Bergmann’s clines in ectotherms: illustrating a life-history perspective with sceloporine lizards. American Naturalist 164: E168–E183. Angilletta, M. J., P. H. Niewiarowski, and C. A. Navas. 2002a. The evolution of thermal physiology in ectotherms. Journal of Thermal Biology 27:249–268.

REFERENCES

Angilletta, M. J., C. E. Ouﬁero, and A. D. Leache. 2006b. Direct and indirect effects of environmental temperature on the evolution of reproductive strategies: An information-theoretic approach. American Naturalist 168:E123-E135. Angilletta, M. J., and M. W. Sears. 2003. Is parental care the key to understanding endothermy in birds and mammals? American Naturalist 162:821–825. Angilletta, M. J., T. D. Steury, and M. W. Sears. 2004a. Temperature, growth rate, and body size in ectotherms: ﬁtting pieces of a life-history puzzle. Integrative and Comparative Biology 44:498–509. Angilletta, M. J., and Y. L. Werner. 1998. Australian geckos do not display diel variation in thermoregulatory behavior. Copeia:736–742. Angilletta, M. J., R. S. Wilson, C. A. Navas, and R. S. James. 2003. Tradeoffs and the evolution of thermal reaction norms. Trends in Ecology & Evolution 18:234–240. Angilletta, M. J., R. S. Winters, and A. E. Dunham. 2000. Thermal effects on the energetics of lizard embryos: implications for hatchling phenotypes. Ecology 81: 2957–2968. Araujo, M. B., and M. Luoto. 2007. The importance of biotic interactions for modelling species distributions under climate change. Global Ecology and Biogeography 16:743–753. Araujo, M. B., R. J. Whittaker, R. J. Ladle, and M. Erhard. 2005. Reducing uncertainty in projections of extinction risk from climate change. Global Ecology and Biogeography 14:529–538. Archer, M. A., J. P. Phelan, K. A. Beckman, and M. R. Rose. 2003. Breakdown in correlations during laboratory evolution. II. Selection on stress resistance in Drosophila populations. Evolution 57:536–543. Armbruster, P., W. E. Bradshaw, K. Ruegg, and C. M. Holzapfel. 2001. Geographic variation and the evolution of reproductive allocation in the pitcher-plant mosquito, Wyeomyia smithii. Evolution 55:439–444. Arnett, A. E., and N. J. Gotelli. 1999. Geographic variation in life-history traits of the ant lion, Myrmeleon immaculatus: evolutionary implications of Bergmann’s rule. Evolution 53:1180–1188. Arnﬁeld, A. J. 2003. Two decades of urban climate research: a review of turbulence, exchanges of energy and water, and the urban heat island. International Journal of Climatology 23:1–26. Arnold, S. J. 1992. Constraints on phenotypic evolution. American Naturalist 140. —. 1994. Multivariate inheritance and evolution: a review of concepts in C. R. B. Boak, ed. Quantitative Genetic Studies of Behavioral Evolution. Chicago, University of Chicago Press.

239

Arnold, S. J., C. R. Peterson, and J. Gladstone. 1995. Behavioral variation in natural populations. VII. Maternal body temperature does not affect juvenile thermoregulation in a garter snake. Animal Behaviour 50:623–633. Artacho, P., L. E. Castaneda, and R. F. Nespolo. 2005. The role of quantitative genetic studies in animal physiological ecology. Revista Chilena De Historia Natural 78:161–167. Asbury, D. A., and S. C. Adolph. 2007. Behavioural plasticity in an ecological generalist: microhabitat use by western fence lizards. Evolutionary Ecology Research 9:801–815. Ashton, K. G. 2002a. Do amphibians follow Bergmann’s rule? Canadian Journal of Zoology 80:708–716. —. 2002b. Patterns of within-species body size variation of birds: strong evidence for Bergmann’s rule. Global Ecology and Biogeography 11:505–523. —. 2004. Sensitivity of intraspeciﬁc latitudinal clines of body size for tetrapods to sampling, latitude and body size. Integrative and Comparative Biology 44:403–412. Ashton, K. G., and C. R. Feldman. 2003. Bergmann’s rule in nonavian reptiles: turtles follow it, lizards and snakes reverse it. Evolution 57:1151–1163. Ashton, K. G., M. C. Tracy, and A. de Queiroz. 2000. Is Bergmann’s rule valid for mammals? American Naturalist 156:390–415. Atkin, O. K., B. R. Loveys, L. J. Atkinson, and T. L. Pons. 2006a. Phenotypic plasticity and growth temperature: understanding interspeciﬁc variability. Journal of Experimental Botany 57:267–281. Atkin, O. K., I. Scheurwater, and T. L. Pons. 2006b. High thermal acclimation potential of both photosynthesis and respiration in two lowland Plantago species in contrast to an alpine congeneric. Global Change Biology 12:500–515. Atkinson, D. 1994. Temperature and organism size: a biological law for ectotherms. Advances in Ecological Research 25:1–58. —. 1996a. Ectotherm life history responses to developmental temperature, Pages 183–204 in I. A. Johnston, and A. F. Bennett, eds. Animals and temperature: phenotypic and evolutionary adaptation. Cambridge, Cambridge University Press. —. 1996b. On the solutions to a major life-history puzzle. Oikos 77:359–365. Atkinson, D., B. J. Ciotti, and D. J. S. Montagnes. 2003. Protists decrease in size linearly with temperature: ca. 2.5% degrees C-1. Proceedings of the Royal Society B 270:2605–2611. Atkinson, D., S. A. Morley, D. Weetman, and R. N. Hughes. 2001. Offspring size responses to maternal temperature in ectotherms, Pages 269–285 in D. Atkinson, and

240

REFERENCES

M. Thorndyke, eds. Environment and Animal Development: Genes, Life Histories and Plasticity. Oxford, Bios Scientiﬁc Publishers. Avery, H. W., J. R. Spotila, J. D. Congdon, R. U. Fischer, E. A. Standora, and S. B. Avery. 1993. Roles of diet protein and temperature in the growth and nutritional energetics of juvenile slider turtles, Trachemys scripta. Physiological Zoology 66:902–925. Ayers, D. Y., and R. Shine. 1997. Thermal inﬂuences on foraging ability: body size, posture and cooling rate of an ambush predator, the python Morelia spilota. Functional Ecology 11:342–347. Ayres, M. P., and J. M. Scriber. 1994. Local adaptation to regional climates in Papilio canadensis (Lepidoptera, Papilionidae). Ecological Monographs 64:465–482. Ayrinhac, A., V. Debat, P. Gibert, A. G. Kister, H. Legout, B. Moreteau, R. Vergilino et al. 2004. Cold adaptation in geographical populations of Drosophila melanogaster: phenotypic plasticity is more important than genetic variability. Functional Ecology 18:700–706. Azevedo, R. B. R., V. French, and L. Partridge. 1996. Thermal evolution of egg size in Drosophila melanogaster. Evolution 50:2338–2345. —. 1997. Life-history consequences of egg size in Drosophila melanogaster. American Naturalist 150:250–282. Baer, C. F., and J. Travis. 2000. Direct and correlated responses to artiﬁcial selection on acute thermal stress tolerance in a livebearing ﬁsh. Evolution 54:238–244. Baird, O. E., and C. C. Krueger. 2003. Behavioral thermoregulation of brook and rainbow trout: comparison of summer habitat use in an Adirondack River, New York. Transactions of the American Fisheries Society 132:1194–1206. Bakken, G. S. 1976. A heat transfer analysis of animals: unifying concepts and the application of metabolism chamber data to ﬁeld ecology. Journal of Theoretical Biology 60:337–384. —. 1981. A two-dimensional operative-temperature model for thermal energy management by animals. Journal of Thermal Biology 6:23–30. —. 1985. Operative and standard operative temperature: tools for thermal and energetic studies. American Zoologist 25:933–943. —. 1989. Arboreal perch properties and the operative temperature experienced by small animals. Ecology 70:922–930. —. 1992. Measurement and application of operative and standard operative temperatures in ecology. American Zoologist 32:194–216. Bakken, G. S., and D. M. Gates. 1975. Heat-transfer analysis of animals: some implications for ﬁeld ecology, physiology, and evolution, Pages 255–290 in D. M. Gates, and

R. B. Schmerl, eds. Perspectives of Biophysical Ecology. New York, Springer-Verlag. Balanya, J., J. M. Oller, R. B. Huey, G. W. Gilchrist, and L. Serra. 2006. Global genetic change tracks global climate warming in Drosophila subobscura. Science 313: 1773–1775. Båmstedt, U., J. Lane, and M. B. Marinussen. 1999. Bioenergetics of ephyra larvae of the scyphozoan jellyﬁsh Aurelia aurita in relation to temperature and salinity. Marine Biology 135. Barber, B. J., and E. C. Crawford. 1977. Stochastic dual-limit hypothesis for behavioral thermoregulation in lizards. Physiological Zoology 50:53–60. Bardoloi, S., and L. K. Hazarika. 1994. Body temperature and thermoregulation of Antheraea assama larvae. Entomologia Experimentalis et Applicata 72:207–217. Barlow, C. A. 1962. The inﬂuence of temperature on the growth of experimental populations of Myzus persicae (Sulzer) and Macrosiphum euphorbiae (Thomas)(Aphididae). Canadian Journal of Zoology 40:145–156. Bartelt, P. E., and C. R. Peterson. 2005. Physically modeling operative temperatures and evaporation rates in amphibians. Journal of Thermal Biology 30:93–102. Bartholomew, G. A. 1981. A matter of size: an examination of endothermy in insects and terrestrial vertebrates, Pages 45–78 in B. Heinrich, ed. Insect Thermoregulation. New York, John Wiley & Sons. —. 1982. Physiological control of body temperature, Pages 167–211 in C. Gans, and F. H. Pough, eds. Biology of the Reptilia. New York, Academic Press. Bartlett, P. N., and D. M. Gates. 1967. Energy budget of a lizard on a tree trunk. Ecology 48:315–322. Basheer, I. A., and M. Hajmeer. 2000. Artiﬁcial neural networks: fundamentals, computing, design, and application. Journal of Microbiological Methods 43:3–31. Bashey, F., and A. E. Dunham. 1997. Elevational variation in the thermal constraints on and microhabitat preferences of the greater earless lizard Cophosaurus texanus. Copeia:725–737. Baur, B., and C. Raboud. 1988. Life history of the land snail Arianta arbustorum along an altitudinal gradient. Journal of Animal Ecology 57:71–87. Bauwens, D., A. M. Castilla, and P. L. N. Mouton. 1999. Field body temperatures, activity levels and opportunities for thermoregulation in an extreme microhabitat specialist, the girdled lizard (Cordylus macropholis). Journal of Zoology 249:11–18. Bauwens, D., T. Garland, A. M. Castilla, and R. Van Damme. 1995. Evolution of sprint speed in lacertid lizards: morphological, physiological, and behavioral covariation. Evolution 49:848–863.

REFERENCES

Bauwens, D., P. E. Hertz, and A. M. Castilla. 1996. Thermoregulation in a lacertid lizard: the relative contributions of distinct behavioral mechanisms. Ecology 77:1818–1830. Bayne, B. L. 1999. Physiological components of growth differences between individual oysters (Crassostrea gigas) and a comparison with Saccostrea commercialis. Physiological and Biochemical Zoology 72:705–713. —. 2000. Relations between variable rates of growth, metabolic costs and growth efﬁciencies in individual Sydney rock oysters (Saccostrea commercialis). Journal of Experimental Marine Biology and Ecology 251:185–203. —. 2004. Phenotypic ﬂexibility and physiological tradeoffs in the feeding and growth of marine bivalve molluscs. Integrative and Comparative Biology 44:425–432. Beaugrand, G., P. C. Reid, F. Ibanez, J. A. Lindley, and M. Edwards. 2002. Reorganization of North Atlantic marine copepod biodiversity and climate. Science 296: 1692–1694. Beaupre, S. J. 1995. Effects of geographically variable thermal environment on bioenergetics of mottled rock rattlesnakes. Ecology 76:1655–1665. Beaupre, S. J., A. E. Dunham, and K. L. Overall. 1993. The effects of consumption rate and temperature on apparent digestibility coefﬁcient, urate production, metabolizable energy coefﬁcient and passage time in canyon lizards (Sceloporus merriami) from two populations. Functional Ecology 7:273–280. Becker, U., G. Colling, P. Dostal, A. Jakobsson, and D. Matthies. 2006. Local adaptation in the monocarpic perennial Carlina vulgaris at different spatial scales across Europe. Oecologia 150:506–518. Beitinger, T. L., W. A. Bennett, and R. W. McCauley. 2000. Temperature tolerances of North American freshwater ﬁshes exposed to dynamic changes in temperature. Environmental Biology of Fishes 58:237–275. Beitinger, T. L., and L. C. Fitzpatrick. 1979. Physiological and ecological correlates of preferred temperature in ﬁsh. American Zoologist 19:319–329. Beitinger, T. L., and J. J. Magnuson. 1975. Inﬂuence of social rank and size on thermoselection behavior of bluegill (Lepomis macrochirus). Journal of the Fisheries Research Board of Canada 32:2133–2136. Belk, M. C., and D. D. Houston. 2002. Bergmann’s rule in ectotherms: a test using freshwater ﬁshes. American Naturalist 160:803–808. Belk, M. C., J. B. Johnson, K. W. Wilson, M. E. Smith, and D. D. Houston. 2005. Variation in intrinsic individual growth rate among populations of leatherside chub (Snyderichthys copei Jordan & Gilbert): adaptation to temperature or length of growing season? Ecology of Freshwater Fish 14:177–184.

241

Bell, G., and S. Collins. 2008. Adaptation, extinction and global change. Evolutionary Applications 1:3–16. Belliure, J., and L. M. Carrascal. 2002. Inﬂuence of heat transmission mode on heating rates and on the selection of patches for heating in a Mediterranean lizard. Physiological and Biochemical Zoology 75:369–376. Bennett, A. F. 1980. The thermal dependence of lizard behaviour. Animal Behaviour 28:752–762. —. 1987a. The accomplishments of ecological physiology, Pages 1–8 in M. E. Feder, A. F. Bennett, R. B. Huey, and W. Burggren, eds. New Directions in Ecological Physiology. New York, Cambridge University Press. —. 1987b. Evolution of the control of body temperature: Is warmer better?, Pages 421–431 in P. Dejours, L. Bolis, C. R. Taylor, and E. R. Weibel, eds. Comparative Physiology: Life in Water and on Land. Padova, Liviana Press. —. 2003. Experimental evolution and the Krogh principle: generating biological novelty for functional and genetic analyses. Physiological and Biochemical Zoology 76: 1–11. Bennett, A. F., J. W. Hicks, and A. J. Cullum. 2000. An experimental test of the thermoregulatory hypothesis for the evolution of endothermy. Evolution 54:1768–1773. Bennett, A. F., and R. E. Lenski. 1993. Evolutionary adaptation to temperature. II. Thermal niches of experimental lines of Escherichia coli. Evolution 47:1–12. —. 1996. Evolutionary adaptation to temperature. V. Adaptive mechanisms and correlated responses in experimental lines of Escherichia coli. Evolution 50: 493–503. —. 1999. Experimental evolution and its role in evolutionary physiology. American Zoologist 39:346–362. Bennett, A. F., R. E. Lenski, and J. E. Mittler. 1992. Evolutionary adaptation to temperature. I. Fitness responses of Escherichia coli to changes in its thermal environment. Evolution 46:16–30. Bennett, A. F., and J. A. Ruben. 1979. Endothermy and activity in vertebrates. Science 206:649–654. Berger, D., R. Walters, and K. Gotthard. 2008. What limits insect fecundity? Body size- and temperaturedependent egg maturation and oviposition in a butterﬂy. Functional Ecology 22:523–529. Berrigan, D. 2000. Correlations between measures of thermal stress resistance within and between species. Oikos 89:301–304. Berrigan, D., and E. L. Charnov. 1994. Reaction norms for age and size at maturity in response to temperature: a puzzle for life historians. Oikos 70:474–478. Berrigan, D., and J. C. Koella. 1994. The evolution of reaction norms: simple models for age and size at maturity. Journal of Evolutionary Biology 7.

242

REFERENCES

Berrigan, D., and S. M. Scheiner. 2004. Modeling the evolution of phenotypic plasticity, Pages 82–97 in T. J. DeWitt, and S. M. Scheiner, eds. Phenotypic Plasticity: Functional and Conceptual Approaches. Oxford, Oxford University Press. Berry, J., and O. Bjorkman. 1980. Photosynthetic response and adaptation to temperature in higher plants. Annual Review of Plant Physiology 31:491–543. Bertalanffy, von, L. 1960. Principles and theory of growth, Pages 137–259 in W. W. Nowinski, ed. Fundamental Aspects of Normal and Malignant Growth. New York, Elsevier. Berven, K. A. 1982. The genetic basis of altitudinal variation in the wood frog Rana sylvatica. II. An experimental analysis of larval development. Oecologia 52: 360–369. Berven, K. A., and D. E. Gill. 1983. Interpreting geographic variation in life-history traits. American Zoologist 23: 85–97. Berven, K. A., D. E. Gill, and S. J. Smith-Gill. 1979. Countergradient selection in the green frog, Rana clamitans. Evolution 33:609–623. Beuchat, C. A., and S. Ellner. 1987. A quantitative test of life-history theory: thermoregulation by a viviparous lizard. Ecological Monographs 57:45–60. Bicego, K. C., R. C. H. Barros, and L. G. S. Branco. 2007. Physiology of temperature regulation: comparative aspects. Comparative Biochemistry and Physiology A 147:616–639. Bilcke, J., S. Downes, and I. Buscher. 2006. Combined effect of incubation and ambient temperature on the feeding performance of a small ectotherm. Austral Ecology 31:937–947. Billerbeck, J. M., T. E. Lankford, and D. O. Conover. 2001. Evolution of intrinsic growth and energy acquisition rates. I. Trade-offs with swimming performance in Menidia menidia. Evolution 55:1863–1872. Billerbeck, J. M., E. T. Schultz, and D. O. Conover. 2000. Adaptive variation in energy acquisition and allocation among latitudinal populations of the Atlantic silverside. Oecologia 122:210–219. Billings, W. D., P. J. Godfrey, B. F. Chabot, and D. P. Bourque. 1971. Metabolic acclimation to temperature in Arctic and Alpine ecotypes of Oxyria digyna. Arctic and Alpine Research 3:277–289. Bishop, J. A., and W. S. Armbruster. 1999. Thermoregulatory abilities of Alaskan bees: effects of size, phylogeny and ecology. Functional Ecology 13:711–724. Black-Samuelsson, S., and S. Andersson. 1997. Reaction norm variation between and within populations of two rare plant species, Vicia pisiformis and V. dumetorum (Fabaceae). Heredity 79:268–276.

Blackburn, T. M., K. J. Gaston, and N. Loder. 1999. Geographic gradients in body size: a clariﬁcation of Bergmann’s rule. Diversity and Distributions 5:165–174. Blanckenhorn, W. U. 2000. Temperature effects on egg size and their ﬁtness consequences in the yellow dung ﬂy Scathophaga stercoraria. Evolutionary Ecology 14: 627–643. Blanckenhorn, W. U., and M. Demont. 2004. Bergmann and converse Bergmann latitudinal clines in arthropods: two ends of a continuum? Integrative and Comparative Biology 44:413–424. Blanckenhorn, W. U., R. C. Stillwell, K. A. Young, C. W. Fox, and K. G. Ashton. 2006. When Rensch meets Bergmann: does sexual size dimorphism change systematically with latitude? Evolution 60:2004–2011. Blázquez, M. C. 1995. Body temperature, activity patterns and movements by gravid and nongravid females of Malpolon monspessulanus. Journal of Herpetology 29: 264–266. Block, B. A., and J. R. Finnerty. 1994. Endothermy in ﬁshes: a phylogenetic analysis of constraints, predispositions, and selection pressures. Environmental Biology of Fishes 40:283–302. Blouin-Demers, G., and P. Nadeau. 2005. The cost–beneﬁt model of thermoregulation does not predict lizard thermoregulatory behavior. Ecology 86:560–566. Blouin-Demers, G., and P. J. Weatherhead. 2001. An experimental test of the link between foraging, habitat selection and thermoregulation in black rat snakes Elaphe obsoleta obsoleta. Journal of Animal Ecology 70:1006–1013. Blows, M. W., and A. A. Hoffman. 2005. A reassessment of genetic limits to evolutionary change. Ecology 86: 1371–1384. Bochdanovits, Z., and G. de Jong. 2003. Experimental evolution in Drosophila melanogaster: interaction of temperature and food quality selection regimes. Evolution 57:1829–1836. Boddy, L. 1983. Effect of temperature and water potential on growth rate of wood-rotting basidiomycetes. Transactions of the British Mycological Society 80: 141–149. Boily, P. 2002. Individual variation in metabolic traits of wild nine-banded armadillos (Dasypus novemcinctus), and the aerobic capacity model for the evolution of endothermy. Journal of Experimental Biology 205: 3207–3214. Bowker, R. G. 1984. Precision of thermoregulation of some African lizards. Physiological Zoology 57:401–412. Bowler, K. 2005. Acclimation, heat shock and hardening. Journal of Thermal Biology 30:125–130. Bradshaw, W. E., and C. M. Holzapfel. 2006. Evolutionary response to rapid climate change. Science 312:1477–1478.

REFERENCES

Braña, F. 1993. Shifts in body temperature and escape behavior of female Podarcis muralis during pregnancy. Oikos 66:216–222. Brattstrom, B. H. 1979. Amphibian temperature regulation studies in the ﬁeld and laboratory. American Zoologist 19:345–356. Brett, J. R., J. E. Shelbourn, and C. T. Shoop. 1969. Growth rate and body composition of ﬁngerling sockeye salmon, Oncorhynchus nerka, in relation to temperature and ration size. Journal of the Fisheries Research Board of Canada 26:2363–2393. Brock, T. D. 1994, Life at high temperatures. Yellowstone National Park, Yellowstone Association for Natural Science. Brodie, E. D., A. J. Moore, and F. J. Janzen. 1995. Visualizing and quantifying natural selection. Trends in Ecology and Evolution 10:313–318. Bronikowski, A. M., A. F. Bennett, and R. E. Lenski. 2001. Evolutionary adaptation to temperature. VII. Effects of temperature on growth rate in natural isolates of Escherichia coli and Salmonella enterica from different thermal environments. Evolution 55:33–40. Brown, G. P., and P. J. Weatherhead. 2000. Thermal ecology and sexual size dimorphism in northern water snakes, Nerodia sipedon. Ecological Monographs 70:311–330. Brown, J. H., J. F. Gillooly, A. P. Allen, V. M. Savage, and G. B. West. 2004. Toward a metabolic theory of ecology. Ecology 85:1771–1789. Brown, R. P., and S. Grifﬁn. 2005. Lower selected body temperatures after food deprivation in the lizard Anolis carolinensis. Journal of Thermal Biology 30: 79–83. Bryant, S. R., and T. G. Shreeve. 2002. The use of artiﬁcial neural networks in ecological analysis: estimating microhabitat temperature. Ecological Entomology 27:424–432. Bryant, S. R., C. D. Thomas, and J. S. Bale. 2000. Thermal ecology of gregarious and solitary nettle-feeding nymphalid butterﬂy larvae. Oecologia 122:1–10. Bubli, O. A., A. G. Imasheva, and V. Loeschcke. 1998. Selection for knockdown resistance to heat in Drosophila melanogaster at high and low larval densities. Evolution 52:619–625. Bubliy, O. A., A. Riihimaa, F. M. Norry, and V. Loeschcke. 2002. Variation in resistance and acclimation to lowtemperature stress among three geographical strains of Drosophila melanogaster. Journal of Thermal Biology 27:337–344. Buchner, O., and G. Neuner. 2003. Variability of heat tolerance in alpine plant species measured at different altitudes. Arctic Antarctic and Alpine Research 35: 411–420.

243

Buckley, L. B. 2008. Linking traits to energetics and population dynamics to predict lizard ranges in changing environments. American Naturalist 171:E1-E19. Buckley, L. B., and J. Roughgarden. 2005. Lizard habitat partitioning on islands: the interaction of local and landscape scales. Journal of Biogeography 32:2113–2121. Bull, J. J., M. R. Badgett, and H. A. Wichman. 2000. Bigbeneﬁt mutations in a bacteriophage inhibited with heat. Molecular Biology and Evolution 17:942–950. Burger, R., and M. Lynch. 1995. Evolution and extinction in a changing environment: a quantitative-genetic analysis. Evolution 49:151–163. Burggren, W. W., and W. E. Bemis. 1990. Studying physiological evolution: paradigms and pitfalls., Pages 191–228 in M. H. Nitecki, ed. Evolutionary Innovations. Chicago, University of Chicago Press. Burnham, K. P., and D. R. Anderson. 2002, Model Selection and Multimodel Inference: A Practical InformationTheoretic Approach. New York, Springer. Campbell, G. S. 1977, An Introduction to Environmental Biophysics. New York, Springer-Verlag. Carlson, S. M., and T. R. Seamons. 2008. A review of quantitative genetic components of ﬁtness in salmonids: implications for adaptation to future change. Evolutionary Applications 1:222–238. Carrière, Y., and G. Boivin. 1997. Evolution of thermal sensitivity of parasitization capacity in egg parasitoids. Evolution 51:2028–2032. —. 2001. Constraints on the evolution of thermal sensitivity of foraging in Trichogramma: genetic trade-offs and plasticity in maternal selection. American Naturalist 157:570–581. Case, T. J., R. D. Holt, M. A. McPeek, and T. H. Keitt. 2005. The community context of species’ borders: ecological and evolutionary perspectives. Oikos 108:28–46. Casey, T. M. 1981. Behavioral mechanisms of thermoregulation, Pages 79–114 in B. Heinrich, ed. Insect Thermoregulation. New York, John Wiley & Sons. —. 1992. Biophysical ecology and heat exchange in insects. American Zoologist 32:225–237. Castañeda, L. E., M. A. Lardies, and F. Bozinovic. 2004. Adaptive latitudinal shifts in the thermal physiology of a terrestrial isopod. Evolutionary Ecology Research 6: 579–593. —. 2005. Interpopulational variation in recovery time from chill coma along a geographic gradient: a study in the common woodlouse, Porcellio laevis. Journal of Insect Physiology 51:1346–1351. Cerda, X., J. Retana, and A. Manzaneda. 1998. The role of competition by dominants and temperature in the foraging of subordinate species in Mediterranean ant communities. Oecologia 117:404–412.

244

REFERENCES

Chamberlin, T. C. 1890. The method of multiple working hypotheses. Science 15:92–97. Chapin, F. S., and M. C. Chapin. 1981. Ecotypic differentiation of growth processes in Carex aquatilis along latitudinal and local gradients. Ecology 62:1000–1009. Chappell, M. A., and D. W. Whitman. 1990. Grasshopper thermoregulation, Pages 43–172 in R. F. Chapman, and A. Joern, eds. Biology of Grasshoppers. New York, Wiley and Sons. Charland, M. B. 1995. Thermal consequences of reptilian viviparity: thermoregulation in gravid and nongravid garter snakes (Thamnophis). Journal of Herpetology 29:383–390. Charland, M. B., and P. T. Gregory. 1990. The inﬂuence of female reproductive status on thermoregulation in a viviparous snake, Crotalus viridis. Copeia:1089–1098. Charnov, E. L., and J. F. Gillooly. 2004. Size and temperature in the evolution of ﬁsh life histories. Integrative and Comparative Biology 44:494–497. Chen, C. P., R. E. Lee, and D. L. Denlinger. 1990. A comparison of the responses of tropical and temperate ﬂies (Diptera, Sarcophagidae) to cold and heat stress. Journal of Comparative Physiology B 160:543–547. Chen, X. J., X. F. Xu, and X. Ji. 2003. Inﬂuence of body temperature on food assimilation and locomotor performance in white-striped grass lizards, Takydromus wolteri (Lacertidae). Journal of Thermal Biology 28:385–391. Chintalapati, S., M. D. Kiran, and S. Shivaji. 2004. Role of membrane lipid fatty acids in cold adaptation. Cellular and Molecular Biology 50:631–642. Chown, S. L., M. D. Le Lagadec, and C. H. Scholtz. 1999. Partitioning variance in a physiological trait: desiccation resistance in keratin beetles (Coleoptera, Trogidae). Functional Ecology 13:838–844. Chown, S. L., and S. W. Nicolson. 2004, Insect Physiological Ecology: Mechanisms and Patterns. Oxford, Oxford University Press. Christian, K., and G. Bedford. 1996. Thermoregulation by the spotted tree monitor, Varanus scalaris, in the seasonal tropics of Australia. Journal of Thermal Biology 21:67–73. Christian, K. A., G. Bedford, B. Green, T. Schultz, and K. Newgrain. 1998. Energetics and water ﬂux of the marbled velvet gecko (Oedura marmorata) in tropical and temperate habitats. Oecologia 116:336–342. Christian, K. A., and G. S. Bedford. 1995. Seasonal changes in thermoregulation by the frillneck lizard, Chlamydosaurus kingii, in tropical Australia. Ecology 76:124–132. Christian, K. A., and C. R. Tracy. 1981. The effect of the thermal environment on the ability of hatchling Galapagos land iguanas to avoid predation during dispersal. Oecologia 49:218–223.

Christian, K. A., C. R. Tracy, and W. P. Porter. 1986. The effect of cold exposure during incubation ofSceloporus undulatus eggs. Copeia 1986:1012–1014. Christian, K. A., and B. W. Weavers. 1996. Thermoregulation of monitor lizards in Australia: an evaluation of methods in thermal biology. Ecological Monographs 66:139–157. Ciais, P., M. Reichstein, N. Viovy, A. Granier, J. Ogee, V. Allard, M. Aubinet et al. 2005. Europe-wide reduction in primary productivity caused by the heat and drought in 2003. Nature 437:529–533. Cichon, M., and J. Kozłowski. 2000. Ageing and typical survivorship curves result from optimal resource allocation. Evolutionary Ecology Research 2:857–870. Clarke, A. 1991. What is cold adaptation and how should we measure it? American Zoologist 31:81–92. Clarke, A., and K. J. Gaston. 2006. Climate, energy and diversity. Proceedings of the Royal Society B 273: 2257–2266. Clarke, A., and P. Rothery. 2008. Scaling of body temperature in mammals and birds. Functional Ecology 22:58–67. Cleavitt, N. 2004. Comparative ecology of a lowland and a subalpine species of Mnium in the northern Rocky Mountains. Plant Ecology 174:205–216. Clusella-Trullas, S., J. S. Terblanche, T. M. Blackburn, and S. L. Chown. 2008. Testing the thermal melanism hypothesis: a macrophysiological approach. Functional Ecology 22:232–238. Coelho, J. R. 2001. Behavioral and physiological thermoregulation in male cicada killers (Sphecius speciosus) during territorial behavior. Journal of Thermal Biology 26:109–116. Cohen, M. P., and R. A. Alford. 1996. Factors affecting diurnal shelter use by the cane toad, Bufo marinus. Herpetologica 52:172–181. Congdon, J. D., and J. W. Gibbons. 1987. Morphological constraint on egg size: a challenge to optimal egg size theory. Proceedings of the National Academy of Sciences 84:4145–4147. Connell, J. H. 1961. The inﬂuence of interspeciﬁc competition and other factors on the distribution of the barnacleChthamalus stellatus. Ecology 42:710–723. Conner, J. K. 2003. Artiﬁcial selection: a powerful tool for ecologists. Ecology 84:1650–1660. Conover, D. O. 1992. Seasonality and the scheduling of life history at different latitudes. Journal of Fish Biology 41:161–178. —. 1998. Local adaptation in marine ﬁshes: evidence and implications for stock enhancement. Bulletin of Marine Science 62:477–493.

REFERENCES

Conover, D. O., and T. M. C. Present. 1990. Countergradient variation in growth rate: compensation for length of the growing season among Atlantic silversides from different latitudes. Oecologia 83:316–324. Conover, D. O., and E. T. Schultz. 1995. Phenotypic similarity and the evolutionary signiﬁcance of countergradient variation. Trends in Ecology & Evolution 10:248–252. Cooper, B. S., B. H. Williams, and M. J. Angilletta. 2008. Unifying indices of heat tolerance in ectotherms. Journal of Thermal Biology 33:in press. Cooper, V. S., A. F. Bennett, and R. E. Lenski. 2001. Evolution of thermal dependence of growth rate of Escherichia coli populations during 20 000 generations in a constant environment. Evolution 55:889–896. Cooper, W. E., and N. Greenberg. 1992. Reptilian coloration and behavior, Pages 298–422 in C. Gans, and D. Crews, eds. Biology of the Reptilia, 18A: Hormones, Brain, and Behavior Chicago, University of Chicago. Cortemeglia, C., and T. L. Beitinger. 2006. Projected US distributions of transgenic and wildtype zebra danios, Danio rerio, based on temperature tolerance data. Journal of Thermal Biology 31:422–428. Cossins, A. R., and K. B. Bowler. 1987, Temperature Biology of Animals. New York, Chapman and Hall. Crill, W. D., R. B. Huey, and G. W. Gilchrist. 1996. Withinand between-generation effects of temperature on the morphology and physiology of Drosophila melanogaster. Evolution 50:1205–1218. Crockett, E. L. 1998. Cholesterol function in plasma membranes from ectotherms: membrane-speciﬁc roles in adaptation to temperature. American Zoologist 38: 291–304. Crompton, A. W., C. R. Taylor, and J. A. Jagger. 1978. Evolution of homeothermy in mammals. Nature 272: 333–336. Crowder, L. B., and J. J. Magnuson. 1983. Cost–beneﬁt analysis of temperature and food resource use: a synthesis with examples from ﬁshes, Pages 189–221 in W. P. Aspey, and S. I. Lustick, eds. Behavioral Energetics. Columbus, Ohio State University Press. Crowley, S. R. 1987. The effect of desiccation upon the preferred body temperature and activity level of the lizard Sceloporus undulatus. Copeia:25–32. Crozier, L., and G. Dwyer. 2006. Combining populationdynamic and ecophysiological models to predict climate-induced insect range shifts. American Naturalist 167:853–866. Cunningham, S., and J. Read. 2003a. Comparison of temperate and tropical rainforest tree species: growth responses to temperature. Journal of Biogeography 30:143–153.

245

Cunningham, S. C., and J. Read. 2002. Comparison of temperate and tropical rainforest tree species: photosynthetic responses to growth temperature. Oecologia 133:112–119. —. 2003b. Do temperate rainforest trees have a greater ability to acclimate to changing temperatures than tropical rainforest trees? New Phytologist 157:55–64. Czarnołeski, M., and J. Kozłowski. 1998. Do Bertalanffy’s growth curves result from optimal resource allocation? Ecology Letters 1:5–7. Danks, H. V. 2004. Seasonal adaptations in arctic insects. Integrative and Comparative Biology 44:85–94. Dark, J. 2005. Annual lipid cycles in hibernators: integration of physiology and behavior. Annual Review of Nutrition 25:469–497. Dark, J., D. R. Miller, P. Licht, and I. Zucker. 1996. Glucoprivation counteracts effects of testosterone on daily torpor in Siberian hamsters. American Journal of Physiology 39:R398-R403. Daut, E. F., and R. M. Andrews. 1993. The effect of pregnancy on thermoregulatory behavior of the viviparous lizard Chalcides ocellatus. Journal of Herpetology 27:6–13. David, J. R., L. O. Araripe, M. Chakir, H. Legout, B. Lemos, G. Petavy, C. Rohmer et al. 2005. Male sterility at extreme temperatures: a signiﬁcant but neglected phenomenon for understanding Drosophila climatic adaptations. Journal of Evolutionary Biology 18:838–846. David, J. R., P. Gibert, E. Gravot, G. Petavy, J. P. Morin, D. Karan, and B. Moreteau. 1997. Phenotypic plasticity and developmental temperature in Drosophila: analysis and signiﬁcance of reaction norms of morphometrical traits. Journal of Thermal Biology 22:441–451. David, J. R., P. Gibert, B. Moreteau, G. W. Gilchrist, and R. B. Huey. 2003. The ﬂy that came in from the cold: geographic variation of recovery time from low-temperature exposure in Drosophila subobscura. Functional Ecology 17:425–430. David, J. R., H. Legout, and B. Moreteau. 2006. Phenotypic plasticity of body size in a temperate population of Drosophila melanogaster: when the temperature-size rule does not apply. Journal of Genetics 85:9–23. Davis, A. J., L. S. Jenkinson, J. H. Lawton, B. Shorrocks, and S. Wood. 1998. Making mistakes when predicting shifts in species range in response to global warming. Nature 391:783–786. Davison, I. R. 1991. Environmental effects on algal photosynthesis: temperature. Journal of Phycology 27:2–8. Dawkins, R. 1980. Good strategy or evolutionarily stable strategy?, Pages 331–367 in G. W. Barlow, and J. Silverberg, eds. Sociobiology: Beyond Nature/Nurture? Boulder, Westview Press.

246

REFERENCES

Dawson, W. R. 1975. On the physiological signiﬁcance of the preferred body temperatures of reptiles, Pages 443–473 in D. M. Gates, and R. B. Schmerl, eds. Perspectives in Biophysical Ecology. Berlin, Springer-Verlag. Day, T. 2000. Competition and the effect of spatial resource heterogeneity on evolutionary diversiﬁcation. American Naturalist 155:790–803. De Block, M., and R. Stoks. 2003. Adaptive sex-speciﬁc life history plasticity to temperature and photoperiod in a damselﬂy. Journal of Evolutionary Biology 16:986–995. De Boer, M. K., E. M. Koolmees, E. G. Vrieling, A. M. Breeman, and M. Van Rijssel. 2005. Temperature responses of three Fibrocapsa japonica strains (Raphidophyceae) from different climate regions. Journal of Plankton Research 27:47–60. de Jong, G. 1995. Phenotypic plasticity as a product of selection in a variable environment. American Naturalist 145:493–512. —. 1999. Unpredictable selection in a structured population leads to local genetic differentiation in evolved reaction norms. Journal of Evolutionary Biology 12: 839–851. de Mazancourt, C., E. Johnson, and T. G. Barraclough. 2008. Biodiversity inhibits species’ evolutionary responses to changing environments. Ecology Letters 11:380–388. de Queiroz, A., and K. G. Ashton. 2004. The phylogeny of a species-level tendency: species heritability and possible deep origins of Bergmann’s rule in tetrapods. Evolution 58:1674–1684. DeAngelis, D. L., and W. M. Mooij. 2005. Individual-based modeling of ecological and evolutionary processes. Annual Review of Ecology Evolution and Systematics 36:147–168. Debat, V., and P. David. 2001. Mapping phenotypes: canalization, plasticity and developmental stability. Trends in Ecology & Evolution 16:555–561. Deere, J. A., and S. L. Chown. 2006. Testing the beneﬁcial acclimation hypothesis and its alternatives for locomotor performance. American Naturalist 168:630–644. Deere, J. A., B. J. Sinclair, D. J. Marshall, and S. L. Chown. 2006. Phenotypic plasticity of thermal tolerances in ﬁve oribatid mite species from sub-Antarctic Marion Island. Journal of Insect Physiology 52:693–700. Denny, M. W., and C. D. G. Harley. 2006. Hot limpets: predicting body temperature in a conductancemediated thermal system. Journal of Experimental Biology 209:2409–2419. Denoel, M., M. Mathieu, and P. Poncin. 2005. Effect of water temperature on the courtship behavior of the Alpine newt Triturus alpestris. Behavioral Ecology and Sociobiology 58:121–127.

Deutsch, C. A., J. J. Tewksbury, R. B. Huey, K. S. Sheldon, C. K. Ghalambor, D. C. Haak, and P. R. Martin. 2008. Impacts of climate warming on terrestrial ectotherms across latitude. Proceedings of the National Academy of Sciences 105:6669–6672. DeWitt, T. J., and R. B. Langerhans. 2004. Integrated solutions to environmental: theory of multimoment reaction norms, Pages 272 in T. J. DeWitt, and S. M. Scheiner, eds. Phenotypic Plasticity: Functional and Conceptual Approaches Oxford, Oxford University Press. DeWitt, T. J., and S. M. Scheiner. 2004. Phenotypic variation from single genotypes: a primer., Pages 1–9 in T. J. DeWitt, and S. M. Scheiner, eds. Phenotypic Plasticity: Functional and Conceptual Approaches. Oxford, Oxford University Press. DeWitt, T. J., A. Sih, and D. S. Wilson. 1998. Costs and limits of phenotypic plasticity. Trends in Ecology & Evolution 13:77–81. Diaz, F., E. Sierra, A. D. Re, and L. Rodriguez. 2002. Behavioural thermoregulation and critical thermal limits of Macrobrachium acanthurus (Wiegman). Journal of Thermal Biology 27:423–428. Diaz, J. A. 1997. Ecological correlates of the thermal quality of an ectotherm’s habitat: a comparison between two temperate lizard populations. Functional Ecology 11:79–89. Diego-Rasilla, F. J. 2003. Inﬂuence of predation pressure on the escape behaviour of Podarcis muralis lizards. Behavioural Processes 63:1–7. Diego-Rasilla, F. J., and V. Perez-Mellado. 2000. The effects of population density on time budgets of the Iberian wall lizard (Podarcis hispanica). Israel Journal of Zoology 46:215–229. Dillon, M. E., M. R. Frazier, and R. Dudley. 2006. Into thin air: physiology and evolution of alpine insects. Integrative and Comparative Biology 46:49–61. Dohm, M. R., W. J. Mautz, P. G. Looby, K. S. Gellert, and J. A. Andrade. 2001. Effects of ozone on evaporative water loss and thermoregulatory behavior of marine toads (Bufo marinus). Environmental Research 86:274–286. Dorcas, M. E., and C. R. Peterson. 1998. Daily body temperature variation in free-ranging rubber boas. Herpetologica 54:88–103. Dorcas, M. E., C. R. Peterson, and M. E. T. Flint. 1997. The thermal biology of digestion in rubber boas (Charina bottae): physiology, behavior, and environmental constraints. Physiological Zoology 70:292–300. Downes, S. 2001. Trading heat and food for safety: costs of predator avoidance in a lizard. Ecology 82:2870–2881. Downes, S., and D. Bauwens. 2002. An experimental demonstration of direct behavioural interference in two

REFERENCES

Mediterranean lacertid lizard species. Animal Behaviour 63:1037–1046. Downes, S., and A. M. Hoefer. 2004. Antipredatory behaviour in lizards: interactions between group size and predation risk. Animal Behaviour 67:485–492. Downes, S., and R. Shine. 1998. Heat, safety or solitude? Using habitat selection experiments to identify a lizard’s priorities. Animal Behaviour 55:1387–1396. Drent, J. 2002. Temperature responses in larvae of Macoma balthica from a northerly and southerly population of the European distribution range. Journal of Experimental Marine Biology and Ecology 275:117–129. Dreyer, B. S., P. Neuenschwander, B. Bouyjou, J. Baumgartner, and S. Dorn. 1997. The inﬂuence of temperature on the life table of Hyperaspis notata. Entomologia Experimentalis Et Applicata 84:85–92. Du, W. G., Y. W. Lu, and J. Y. Shen. 2005. The inﬂuence of maternal thermal environments on reproductive traits and hatchling traits in a lacertid lizard, Takydromus septentrionalis. Journal of Thermal Biology 30:153–161. Du, W. G., S. J. Yan, and X. Ji. 2000. Selected body temperature, thermal tolerance and thermal dependence of food assimilation and locomotor performance in adult blue-tailed skinks, Eumeces elegans. Journal of Thermal Biology 25:197–202. Duellman, W. E., and L. Trueb. 1986, Biology of Amphibians. New York, McGraw Hill. Dufresne, F., and P. D. N. Hebert. 1998. Temperaturerelated differences in life-history characteristics between diploid and polyploid clones of the Daphnia pulex complex. Ecoscience 5:433–437. Dulai, S., I. Molnar, and E. Lehoczki. 1998. Effects of growth temperatures of 5 and 25 degrees C on long-term responses of photosystem II to heat stress in atrazineresistant and susceptible biotypes of Erigeron canadensis. Australian Journal of Plant Physiology 25:145–153. Dullinger, S., T. Dirnbock, and G. Grabherr. 2004. Modelling climate change-driven treeline shifts: relative effects of temperature increase, dispersal and invasibility. Journal of Ecology 92:241–252. Dunham, A. E., B. W. Grant, and K. L. Overall. 1989. Interfaces between biophysical and physiological ecology and the population ecology of terrestrial vertebrate ectotherms. Physiological Zoology 62:335–355. Dutton, R. H., L. C. Fitzpatrick, and J. L. Hughes. 1975. Energetics of the rusty lizard Sceloporus olivaceus. Ecology 56:1378–1387. Dwyer, S. A., O. Ghannoum, A. Nicotra, and S. Von Caemmerer. 2007. High temperature acclimation of C-4 photosynthesis is linked to changes in photosynthetic biochemistry. Plant Cell and Environment 30: 53–66.

247

Dybdahl, M. F., and S. L. Kane. 2005. Adaptation vs. phenotypic plasticity in the success of a clonal invader. Ecology 86:1592–1601. Dzialowski, E. M. 2005. Use of operative temperature and standard operative temperature models in thermal biology. Journal of Thermal Biology 30:317–334. Dzialowski, E. M., and M. P. O’Connor. 2001. Physiological control of warming and cooling during simulated shuttling and basking in lizards. Physiological and Biochemical Zoology 74:679–693. —. 2004. Importance of the limbs in the physiological control of heat exchange in Iguana iguana and Sceloporus undulatus. Journal of Thermal Biology 29:299–305. Edmunds, P. J. 2005. The effect of sub-lethal increases in temperature on the growth and population trajectories of three scleractinian corals on the southern Great Barrier Reef. Oecologia 146:350–364. Eggert, A., E. M. Burger, and A. M. Breeman. 2003a. Ecotypic differentiation in thermal traits in the tropical to warm-temperate green macrophyte Valonia utricularis. Botanica Marina 46:69–81. Eggert, A., P. R. Van Hasselt, and A. M. Breeman. 2003b. Differences in thermal acclimation of chloroplast functioning in two ecotypes of Valonia utricularis (Chlorophyta). European Journal of Phycology 38:123–131. Eggert, A., R. J. W. Visser, P. R. Van Hasselt, and A. M. Breeman. 2006. Differences in acclimation potential of photosynthesis in seven isolates of the tropical to warm temperate macrophyteValonia utricularis (Chlorophyta). Phycologia 45:546–556. Eggert, A., and C. Wiencke. 2000. Adaptation and acclimation of growth and photosynthesis of ﬁve Antarctic red algae to low temperatures. Polar Biology 23:609–618. Ehleringer, J., and I. Forseth. 1980. Solar tracking by plants. Science 210:1094–1098. Elith, J., C. H. Graham, R. P. Anderson, M. Dudik, S. Ferrier, A. Guisan, R. J. Hijmans et al. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29:129–151. Elliott, J. M. 1982. The effects of temperature and ration size on the growth and energetics of salmonids in captivity. Comparative Biochemistry and Physiology B 73:81–91. Ellison, A. M. 2004. Bayesian inference in ecology. Ecology Letters 7:509–520. Ellison, G. T. H., and J. D. Skinner. 1992. The inﬂuence of ambient temperature on spontaneous daily torpor in pouched mice (Saccostomus campestris: Rodentia— Cricetidae) from southern Africa. Journal of Thermal Biology 17:25–31. Else, P. L., N. Turner, and A. J. Hulbert. 2004. The evolution of endothermy: role for membranes and molecular activity. Physiological and Biochemical Zoology 77:950–958.

248

REFERENCES

Else, P. L., and B. J. Wu. 1999. What role for membranes in determining the higher sodium pump molecular activity of mammals compared to ectotherms? Journal of Comparative Physiology B 169:296–302. Endler, J. A. 1986, Natural Selection in the Wild. Princeton, Princeton University Press. Eppley, R. W. 1972. Temperature and phytoplankton growth in sea. Fishery Bulletin 70:1063–1085. Falconer, D. S. 1952. The problem of environment and selection. American Naturalist 86:293–298. —. 1989, Introduction to Quantitative Genetics. New York, John Wiley & Sons, Inc. Fangue, N. A., M. Hofmeister, and P. M. Schulte. 2006. Intraspeciﬁc variation in thermal tolerance and heat shock protein gene expression in common killiﬁsh, Fundulus heteroclitus. Journal of Experimental Biology 209:2859–2872. Farmer, C. G. 2000. Parental care: the key to understanding endothermy and other convergent features in birds and mammals. American Naturalist 155:326–334. Feder, M. E., A. F. Bennett, and R. B. Huey. 2000. Evolutionary physiology. Annual Review of Ecology and Systematics 31:315–341. Feder, M. E., and G. E. Hofmann. 1999. Heat-shock proteins, molecular chaperones, and the stress response: evolutionary and ecological physiology. Annual Review of Physiology 61:243–282. Feder, M. E., and J. F. Lynch. 1982. Effects of latitude, season, elevation, and microhabitat on ﬁeld body temperatures of neotropical and temperate zone salamanders. Ecology 63:1657–1664. Felsenstein, J. 1985. Phylogenies and the comparative method. American Naturalist 125:1–15. —. 2002. Contrasts for a within-species comparative method, Pages 118–129 in M. Slatkin, and M. Veuille, eds. Modern Developments in Theoretical Population Genetics. Oxford, Oxford University Press. Fields, P. A. 2001. Protein function at thermal extremes: balancing stability and ﬂexibility. Comparative Biochemistry and Physiology A 129:417–431. Fischer, K., A. N. M. Bot, P. M. Brakeﬁeld, and B. J. Zwaan. 2003c. Fitness consequences of temperature-mediated egg size plasticity in a butterﬂy. Functional Ecology 17:803–810. Fischer, K., P. M. Brakeﬁeld, and B. J. Zwaan. 2003b. Plasticity in butterﬂy egg size: why larger offspring at lower temperatures? Ecology 84:3138–3147. Fischer, K., E. Eenhoorn, A. N. M. Bot, P. M. Brakeﬁeld, and B. J. Zwaan. 2003a. Cooler butterﬂies lay larger eggs: developmental plasticity versus acclimation. Proceedings of the Royal Society B 270:2051–2056.

Fischer, K., and K. Fiedler. 2000. Sex-related differences in reaction norms in the butterﬂy Lycaena tityrus (Lepidoptera: Lycaenidae). Oikos 90:372–380. —. 2001a. Dimorphic growth patterns and sex-speciﬁc reaction norms in the butterﬂy Lycaena hippothoe sumadiensis. Journal of Evolutionary Biology 14:210–218. —. 2001b. Egg weight variation in the butterﬂy Lycaena hippothoe: more small or fewer large eggs? Population Ecology 43:105–109. —. 2002. Life-history plasticity in the butterﬂy Lycaena hippothoe: local adaptations and trade-offs. Biological Journal of the Linnean Society 75:173–185. Fischer, K., B. J. Zwaan, and P. M. Brakeﬁeld. 2004. Genetic and environmental sources of egg size variation in the butterﬂy Bicyclus anynana. Heredity 92:163–169. Fisher, R. A. 1930, The Genetical Theory of Natural Selection. Oxford, Oxford University Press. Fitter, A. H., and R. S. R. Fitter. 2002. Rapid changes in ﬂowering time in British plants. Science 296: 1689–1691. Fleming, I. A., and M. R. Gross. 1990. Latitudinal clines: a trade-off between egg number and size in Paciﬁc salmon. Ecology 71:1–11. Folk, D. G., P. Zwollo, D. M. Rand, and G. W. Gilchrist. 2006. Selection on knockdown performance in Drosophila melanogaster impacts thermotolerance and heat-shock response differently in females and males. Journal of Experimental Biology 209:3964–3973. Forchhammer, M. C., and E. Post. 2000. Climatic signatures in ecology. Trends in Ecology & Evolution 15: 286–286. Forseth, I., and J. R. Ehleringer. 1980. Solar tracking response to drought in a desert annual. Oecologia 44:159–163. Fox, C. W., and M. E. Czesak. 2000. Evolutionary ecology of progeny size in arthropods. Annual Review of Entomology 45:341–369. Franco, A. M. A., J. K. Hill, C. Kitschke, Y. C. Collingham, D. B. Roy, R. Fox, B. Huntley et al. 2006. Impacts of climate warming and habitat loss on extinctions at species’ low-latitude range boundaries. Global Change Biology 12:1545–1553. Frazier, M., R. B. Huey, and D. Berrigan. 2006. Thermodynamics constrains the evolution of insect population growth rate: “warmer is better”. American Naturalist 168:512–520. Freidenburg, L. K., and D. K. Skelly. 2004. Microgeographical variation in thermal preference by an amphibian. Ecology Letters 7:369–373. Fretwell, S. D., and H. L. Lucas. 1969. On theoretical behavior and other factors inﬂuencing habitat distribution in birds. Acta Biotheoretica 19.

REFERENCES

Frid, L., and J. H. Myers. 2002. Thermal ecology of western tent caterpillars Malacosoma californicum pluviale and infection by nucleopolyhedrovirus. Ecological Entomology 27:665–673. Fry, F. E. J., and J. S. Hart. 1948. Cruising speed of goldﬁsh in relation to water temperature Journal of the Fisheries Research Board of Canada 7:169–175. Full, R. J., and A. Tullis. 1990. Capacity for sustained terrestrial locomotion in an insect: energetics, thermal dependence, and kinematics. Journal of Comparative Physiology B 160:573–581. Gabriel, W. 1988. Quantitative genetic models for parthenogenetic species, Pages 73–82 in G. de Jong, ed. Population Genetics and Evolution. Berlin, SpringerVerlag. —. 1999. Evolution of reversible plastic responses: inducible defenses and environmental tolerance, Pages 286–305 in C. D. Harvell, and R. Tollrian, eds. The Ecology and Evolution of Inducible Defenses. Princeton, Princeton University Press. —. 2005. How stress selects for reversible phenotypic plasticity. Journal of Evolutionary Biology 18:873–883. —. 2006. Selective advantage of irreversible and reversible phenotypic plasticity. Archiv Fur Hydrobiologie 167: 1–20. Gabriel, W., B. Luttbeg, A. Sih, and R. Tollrian. 2005. Environmental tolerance, heterogeneity, and the evolution of reversible plastic responses. American Naturalist 166:339–353. Gabriel, W., and M. Lynch. 1992. The selective advantage of reaction norms for environmental tolerance. Journal of Evolutionary Biology 5:41–59. Gadgil, M., and W. H. Bossert. 1970. Life historical consequences of natural selection. American Naturalist 104:1–24. Gagliano, M., M. I. McCormick, and M. G. Meekan. 2007. Temperature-induced shifts in selective pressure at a critical developmental transition. Oecologia 152:219–225. Galen, C. 2006. Solar furnaces or swamp coolers: costs and beneﬁts of water use by solar-tracking ﬂowers of the alpine snow buttercup, Ranunculus adoneus. Oecologia 148:195–201. Garbuz, D., M. B. Evgenev, M. E. Feder, and O. G. Zatsepina. 2003. Evolution of thermotolerance and the heat-shock response: evidence from inter/intraspeciﬁc comparison and interspeciﬁc hybridization in the virilis species group of Drosophila. I. Thermal phenotype. Journal of Experimental Biology 206:2399–2408. Garland, T., and S. C. Adolph. 1994. Why not to do twospecies comparative studies: limitations on inferring adaptation. Physiological Zoology 67:797–828.

249

Garland, T., A. F. Bennett, and E. L. Rezende. 2005. Phylogenetic approaches in comparative physiology. Journal of Experimental Biology 208:3015–3035. Garland, T., and P. A. Carter. 1994. Evolutionary physiology. Annual Review of Physiology 56:579–621. Garland, T., P. H. Harvey, and A. R. Ives. 1992. Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology 41:18–32. Garland, T., R. B. Huey, and A. F. Bennett. 1991. Phylogeny and coadaptation of thermal physiology in lizards: a reanalysis. Evolution 45:1969–1975. Garland, T., P. E. Midford, and A. R. Ives. 1999. An introduction to phylogenetically based statistical methods, with a new method for conﬁdence intervals on ancestral values. American Zoologist 39:374–388. Gaston, K. J., and S. L. Chown. 1999. Elevation and climatic tolerance: a test using dung beetles. Oikos 86:584–590. Gaston, K. J., and J. I. Spicer. 1998. Do upper thermal tolerances differ in geographically separated populations of the beachﬂea Orchestia gammarellus (Crustacea: Amphipoda)? Journal of Experimental Marine Biology and Ecology 229:265–276. Gates, D. M. 1980, Biophysical Ecology. New York, Springer-Verlag. Gavrilets, S. 1997. Coevolutionary chase in exploitervictim systems with polygenic characters. Journal of Theoretical Biology 186:527–534. Gavrilets, S., and S. M. Scheiner. 1993. The genetics of phenotypic plasticity. VI. Theoretical predictions for directional selection. Journal of Evolutionary Biology 6:49–68. Gebhardt, M. D., and S. C. Stearns. 1993. Phenotypic plasticity for life-history traits in Drosophila melanogaster. I. Effect on phenotypic and environmental correlations. Journal of Evolutionary Biology 6:1–16. Geiser, F. 1998. Evolution of daily torpor and hibernation in birds and mammals: importance of body size. Clinical and Experimental Pharmacology and Physiology 25:736–739. Geiser, F., G. Kortner, and I. Schmidt. 1998. Leptin increases energy expenditure of a marsupial by inhibition of daily torpor. American Journal of Physiology 44:R1627–R1632. Gerald, G. W., and L. C. Spezzano. 2005. The inﬂuence of chemical cues and conspeciﬁc density on the temperature selection of a freshwater snail (Melanoides tuberculata). Journal of Thermal Biology 30:237–245. Ghalambor, C. K., R. B. Huey, P. R. Martin, J. J. Tewksbury, and G. Wang. 2006. Are mountain passes higher in the tropics? Janzen’s hypothesis revisited. Integrative and Comparative Biology 46:5–17.

250

REFERENCES

Gibert, P., and R. B. Huey. 2001. Chill-coma temperature in Drosophila: Effects of developmental temperature, latitude, and phylogeny. Physiological and Biochemical Zoology 74:429–434. Gibert, P., B. Moreteau, J. R. David, and S. M. Scheiner. 1998. Describing the evolution of reaction norm shape: Body pigmentation in Drosophila. Evolution 52: 1501–1506. Gibert, P., B. Moreteau, G. Petavy, D. Karan, and J. R. David. 2001. Chill-coma tolerance, a major climatic adaptation among Drosophila species. Evolution 55: 1063–1068. Gilchrist, A. S., R. B. R. Azevedo, L. Partridge, and P. O’Higgins. 2000. Adaptation and constraint in the evolution of Drosophila melanogaster wing shape. Evolution & Development 2:114–124. Gilchrist, G. W. 1995. Specialists and generalists in changing environments. I. Fitness landscapes of thermal sensitivity. American Naturalist 146:252–270. —. 1996. A quantitative genetic analysis of thermal sensitivity in the locomotor performance curve of Aphidius ervi. Evolution 50:1560–1572. —. 2000. The evolution of thermal sensitivity in changing environments in K. B. Storey, and J. M. Storey, eds. Cell and Molecular Responses to Stress. Volume 1. Environmental Stressors and Gene Responses. Amsterdam, Elsevier Science. Gilchrist, G. W., and R. B. Huey. 1999. The direct response of Drosophila melanogaster to selection on knockdown temperature. Heredity 83:15–29. —. 2001. Parental and developmental temperature effects on the thermal dependence of ﬁtness in Drosophila melanogaster. Evolution 55:209–214. —. 2004. Plastic and genetic variation in wing loading as a function of temperature within and among parallel clines in Drosophila subobscura. Integrative and Comparative Biology 44:461–470. Gilchrist, G. W., R. B. Huey, J. Balanya, M. Pascual, and L. Serra. 2004. A time series of evolution in action: a latitudinal cline in wing size in South American Drosophila subobscura. Evolution 58:768–780. Gilchrist, G. W., R. B. Huey, and L. Partridge. 1997. Thermal sensitivity of Drosophila melanogaster: evolutionary responses of adults and eggs to laboratory natural selection at different temperatures. Physiological Zoology 70:403–414. Gilchrist, G. W., and J. G. Kingsolver. 2001. Is optimality over the hill?, Pages 219–241 in S. H. Orzack, and E. Sober, eds. Adaptationism and Optimality. Cambridge, Cambridge University Press. Gillespie, J. 2004, Population Genetics: A Concise Guide. Baltimore, Johns Hopkins University Press.

Gilligan, M. R. 1991. Bergmann ecogeographic trends among the tripleﬁn blennies (Teleostei: Tripterygiidae) in the Gulf of California, Mexico. Environmental Biology of Fishes 31:301–305. Gillis, R. 1991. Thermal biology of two populations of red-chinned lizards (Sceloporus undulatus erythrocheilus) living in different habitats in southcentral Colorado. Journal of Herpetology 25:18–23. Gillooly, J. F., J. H. Brown, G. B. West, V. M. Savage, and E. L. Charnov. 2001. Effects of size and temperature on metabolic rate. Science 293:2248–2251. Gillooly, J. F., E. L. Charnov, G. B. West, V. M. Savage, and J. H. Brown. 2002. Effects of size and temperature on developmental time. Nature 417:70–73. Gilman, S. E. 2006. Life at the edge: an experimental study of a poleward range boundary. Oecologia 148: 270–279. Gilman, S. E., D. S. Wethey, and B. Helmuth. 2006. Variation in the sensitivity of organismal body temperature to climate change over local and geographic scales. Proceedings of the National Academy of Sciences 103:9560–9565. Gittleman, J. L., and H. K. Luh. 1994. Phylogeny, evolutionary models and comparative methods: a simulation study, Pages 103–122 in E. P., and R. Vane-Wright, eds. Phylogenetics and Ecology. London, Academic Press. Glanville, E. J., and F. Seebacher. 2006. Compensation for environmental change by complementary shifts of thermal sensitivity and thermoregulatory behaviour in an ectotherm. Journal of Experimental Biology 209: 4869–4877. Gomes, F. R., J. G. Chaui-Berlinck, J. Bicudo, and C. A. Navas. 2004. Intraspeciﬁc relationships between resting and activity metabolism in anuran amphibians: inﬂuence of ecology and behavior. Physiological and Biochemical Zoology 77:197–208. Gomi, T., M. Nagasaka, T. Fukuda, and H. Hagihara. 2007. Shifting of the life cycle and life-history traits of the fall webworm in relation to climate change. Entomologia Experimentalis et Applicata 125:179–184. Gomulkiewicz, R., and M. Kirkpatrick. 1992. Quantitative genetics and the evolution of reaction norms. Evolution 46:390–411. Good, D. S. 1993. Evolution of behaviours in Drosophila melanogaster in high temperatures: genetic and environmental effects. Journal of Insect Physiology 39:537–544. Gordo, O., and J. J. Sanz. 2005. Phenology and climate change: a long-term study in a Mediterranean locality. Oecologia 146:484–495. Gotthard, K. 2000. Increased risk of predation as a cost of high growth rate: an experimental test in a butterﬂy. Journal of Animal Ecology 69:896–902.

REFERENCES

Gotthard, K., and S. Nylin. 1995. Adaptive plasticity and plasticity as an adaptation: a selective review of plasticity in animal morphology and life history. Oikos 74. Graham, C. H., S. Ferrier, F. Huettman, C. Moritz, and A. T. Peterson. 2004. New developments in museumbased informatics and applications in biodiversity analysis. Trends in Ecology & Evolution 19:497–503. Grant, B. W. 1990. Trade-offs in activity time and physiological performance for thermoregulating desert lizards, Sceloporus merriami. Ecology 71:2323–2333. Grant, B. W., and A. E. Dunham. 1988. Thermally imposed time constraints on the activity of the desert lizard Sceloporus merriami. Ecology 69:167–176. —. 1990. Elevational covariation in environmental constraints and life histories of the desert lizard Sceloporus merriami. Ecology 71:1765–1776. Grant, B. W., and W. P. Porter. 1992. Modeling global macroclimatic constraints on ectotherm energy budgets. American Zoologist 32:154–178. Gratani, L., and L. Varone. 2004. Adaptive photosynthetic strategies of the Mediterranean maquis species according to their origin. Photosynthetica 42:551–558. Graves, B. M., and D. Duvall. 1993. Reproduction, rookery use, and thermoregulation in free-ranging, pregnant Crotalus v. viridis. Journal of Herpetology 27:33–41. Greenwald, O. E. 1974. Thermal dependence of striking and prey capture by gopher snakes. Copeia 1974: 141–148. Gregory, P. T., L. H. Crampton, and K. M. Skebo. 1999. Conﬂicts and interactions among reproduction, thermoregulation and feeding in viviparous reptiles: are gravid snakes anorexic? Journal of Zoology 248:231–241. Grether, G. F. 2005. Environmental change, phenotypic plasticity, and genetic compensation. American Naturalist 166:E115-E123. Grewal, P. S., S. Selvan, and R. Gaugler. 1994. Thermal adaptation of entomopathogenic nematodes: niche breadth for infection, establishment, and reproduction. Journal of Thermal Biology 19:245–253. Grifﬁths, J. S., and D. F. Alderdice. 1972. Effects of acclimation and acute temperature experience on swimming speed of juvenile coho salmon. Journal of the Fisheries Research Board of Canada 29:251–264. Grigg, G. C. 2004. An evolutionary framework for studies of hibernation and short-term torpor Pages 131–141 in B. M. Barnes, and H. V. Carey, eds. Life in the Cold: Evolution, Adaptation, Mechanisms and Applications. Twelfth International Hibernation Symposium, Biological Papers of the University of Alaska. Grigg, G. C., and L. Beard. 2000. Hibernation by echidnas in mild climates: hints about the evolution of endothermy?, Pages 5–19 in G. Heldmaier, M. Klingenspor, and A. P.

251

Beard, eds. Life In the Cold: Eleventh International Hibernation Symposium. Berlin, Springer. Grigg, G. C., L. A. Beard, and M. L. Augee. 2004. The evolution of endothermy and its diversity in mammals and birds. Physiological and Biochemical Zoology 77:982–997. Grigg, G. C., and F. Seebacher. 1999. Field test of a paradigm: hysteresis of heart rate in thermoregulation by a free-ranging lizard (Pogona barbata). Proceedings of the Royal Society B 266:1291–1297. Grimm, N. B., S. H. Faeth, N. E. Golubiewski, C. L. Redman, J. G. Wu, X. M. Bai, and J. M. Briggs. 2008. Global change and the ecology of cities. Science 319: 756–760. Guderley, H. 2004. Metabolic responses to low temperature in ﬁsh muscle. Biological Reviews 79:409–427. Guderley, H., P. H. Leroy, and A. Gagne. 2001. Thermal acclimation, growth, and burst swimming of threespine stickleback: enzymatic correlates and inﬂuence of photoperiod. Physiological and Biochemical Zoology 74:66–74. Guisan, A., and W. Thuiller. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8:993–1009. Gunderson, C. A., R. J. Norby, and S. D. Wullschleger. 2000. Acclimation of photosynthesis and respiration to simulated climatic warming in northern and southern populations of Acer saccharum: laboratory and ﬁeld evidence. Tree Physiology 20:87–96. Gurevitch, J., and L. V. Hedges. 1993. Meta-analysis: combining the results of independent experiments, Pages 378–398 in S. Scheiner, and J. Gurevitch, eds. Design and analysis of experiments. New York, Chapman and Hall. Gvoždík, L. 2002. To heat or to save time? Thermoregulation in the lizard Zootoca vivipara (Squamata: Lacertidae) in different thermal environments along an altitudinal gradient. Canadian Journal of Zoology 80:479–492. —. 2003. Postprandial thermophily in the Danube crested newt, Triturus dobrogicus. Journal of Thermal Biology 28:545–550. Gvoždík, L., and A. M. Castilla. 2001. A comparative study of preferred body temperatures and critical thermal tolerance limits among populations of Zootoca vivipara (Squamata: Lacertidae) along an altitudinal gradient. Journal of Herpetology 35:486–492. Gvoždík, L., M. Puky, and M. Sugerkova. 2007. Acclimation is beneﬁcial at extreme test temperatures in the Danube crested newt, Triturus dobrogicus (Caudata, Salamandridae). Biological Journal of the Linnean Society 90:627–636. Hahn, M. W., and M. Pockl. 2005. Ecotypes of planktonic actinobacteria with identical 16S rRNA genes adapted

252

REFERENCES

to thermal niches in temperate, subtropical, and tropical freshwater habitats. Applied and Environmental Microbiology 71:766–773. Haldimann, P., and U. Feller. 2005. Growth at moderately elevated temperature alters the physiological response of the photosynthetic apparatus to heat stress in pea (Pisum sativum L.) leaves. Plant Cell and Environment 28:302–317. Hallas, R., M. Schiffer, and A. A. Hoffmann. 2002. Clinal variation in Drosophila serrata for stress resistance and body size. Genetical Research 79:141–148. Hansen, T. F. 2006. The evolution of genetic architecture. Annual Review of Ecology, Evolution, and Systematics 37:123–157. Harrington, R., I. Woiwod, and T. Sparks. 1999. Climate change and trophic interactions. Trends in Ecology & Evolution 14:146–150. Harrison, J. F., J. H. Fewell, S. P. Roberts, and H. G. Hall. 1996. Achievement of thermal stability by varying metabolic heat production in ﬂying honeybees. Science 274:88–90. Harvell, C. D., C. E. Mitchell, J. R. Ward, S. Altizer, A. P. Dobson, R. S. Ostfeld, and M. D. Samuel. 2002. Climate warming and disease risks for terrestrial and marine biota. Science 296:2158–2162. Harvey, P. H., and A. Purvis. 1991. Comparative methods for explaining adaptations. Nature 351:619–624. Harwood, R. H. 1979. Effect of temperature on the digestive efﬁciency of 3 species of lizards, Cnemidophorus tigris, Gerrhonotus multicarinatus and Sceloporus occidentalis. Comparative Biochemistry and Physiology A 63:417–433. Hawkins, A. J. S. 1991. Protein turnover: a functional appraisal. Functional Ecology 5:222–233. —. 1995. Effects of temperature change on ectotherm metabolism and evolution: metabolic and physiological interrelations underlying the superiority of multilocus heterozygotes in heterogeneous environments. Journal of Thermal Biology 20:23–33. Hayes, J. P., and T. Garland. 1995. The evolution of endothermy: testing the aerobic capacity model. Evolution 49:836–847. Haynie, D. T. 2001, Biological Thermodynamics. Cambridge, Cambridge University Press. Hazel, J. R. 1995. Thermal adaptation in biological membranes: is homeoviscous adaptation the explanation? Annual Review of Physiology 57:19–42. Hazel, J. R., and E. E. Williams. 1990. The role of alterations in membrane lipid composition in enabling physiological adaptation of organisms to their physical environment. Progress in Lipid Research 29:167–227.

Hazel, W. N. 2002. The environmental and genetic control of seasonal polyphenism in larval color and its adaptive signiﬁcance in a swallowtail butterﬂy. Evolution 56: 342–348. Heath, J. E. 1964. Reptilian thermoregulation: evaluation of ﬁeld studies. Science 146:784–785. —. 1965. Temperature regulation and diurnal activity in horned lizards. University of California Publications in Zoology 64:97–136. Heckrotte, C. 1983. The inﬂuence of temperature on the behavior of the Mexican jumping bean. Journal of Thermal Biology 8:333–335. Heino, M., and V. Kaitala. 1999. Evolution of resource allocation between growth and reproduction in animals with indeterminate growth. Journal of Evolutionary Biology 12:423–429. Heinrich, B. 1974. Thermoregulation in endothermic insects. Science 185:747–756. —. 1981. Insect Thermoregulation. New York, John Wiley & Sons. —. 1990. Is reﬂectance basking real? Journal of Experimental Biology 154:31–43. Helmuth, B. S. T. 1998. Intertidal mussel microclimates: predicting the body temperature of a sessile invertebrate. Ecological Monographs 68:51–74. Helmuth, B. S. T., and G. E. Hofmann. 2001. Microhabitats, thermal heterogeneity, and patterns of physiological stress in the rocky intertidal zone. Biological Bulletin 201:374–384. Hendry, A. P., T. Day, and A. B. Cooper. 2001. Optimal size and number of propagules: allowance for discrete stages and effects of maternal size on reproductive output and offspring ﬁtness. American Naturalist 157:387–407. Henwood, K. 1975. A ﬁeld-tested thermoregulation model for two diurnal Namib desert tenebrionid beetles. Ecology 56:1329–1342. Hercus, M. J., V. Loeschcke, and S. I. S. Rattan. 2003. Lifespan extension of Drosophila melanogaster through hormesis by repeated mild heat stress. Biogerontology 4:149–156. Herczeg, G., A. Gonda, J. Saarikivi, and J. Merila. 2006. Experimental support for the cost–beneﬁt model of lizard thermoregulation. Behavioral Ecology and Sociobiology 60:405–414. Herczeg, G., A. Herrero, J. Saarikivi, A. Gonda, M. Jantti, and J. Merila. 2008. Experimental support for the cost– beneﬁt model of lizard thermoregulation: the effects of predation risk and food supply. Oecologia 155:1–10. Hereford, J., T. F. Hansen, and D. Houle. 2004. Comparing strengths of directional selection: How strong is strong? Evolution 58:2133–2143.

REFERENCES

Hernandez, R. M., R. F. Buckle, E. Palacios, and S. B. Baron. 2006. Preferential behavior of white shrimp Litopenaeus vannamei (Boone 1931) by progressive temperaturesalinity simultaneous interaction. Journal of Thermal Biology 31:565–572. Hertz, A., and D. Kobler. 2000. A framework for the description of evolutionary algorithms. European Journal of Operational Research 126:1–12. Hertz, P. E. 1974. Thermal passivity of a tropical forest lizard, Anolis polylepis. Journal of Herpetology 8:323–327. —. 1979. Sensitivity to high temperatures in three West Indian grass anoles (Sauria, Iguanidae), with a review of heat sensitivity in the genus Anolis. Comparative Biochemistry and Physiology A 63:217–222. —. 1981. Adaptation to altitude in two West Indian anoles (Reptilia, Iguanidae): ﬁeld thermal biology and physiological ecology. Journal of Zoology 195:25–37. —. 1983. Eurythermy and niche breadth in West Indian Anolis lizards: a reappraisal, Pages 472–483 in G. J. Rhodin, and K. I. Miyata, eds. Advances in Herpetology and Evolutionary: Essays in Honor of Ernest E. Williams. Cambridge, Museum of Comparative Zoology, Harvard University. —. 1992. Temperature regulation in Puerto Rican Anolis lizards: a ﬁeld test using null hypotheses. Ecology 73:1405–1417. Hertz, P. E., A. Arce-Hernandez, J. Ramirez-Vazquez, W. Tirado-Rivera, and L. Vazquez-Vives. 1979. Geographical variation of heat sensitivity and water loss rates in the tropical lizard, Anolis gundlachi. Comparative Biochemistry and Physiology A 62:947–953. Hertz, P. E., and R. B. Huey. 1981. Compensation for altitudinal changes in the thermal environment by some Anolis lizards on Hispaniola. Ecology 62:515–521. Hertz, P. E., R. B. Huey, and T. Garland. 1988. Time budgets, thermoregulation, and maximal locomotor performance: are reptiles Olympians or boy scouts? American Zoologist 28:927–938. Hertz, P. E., R. B. Huey, and E. Nevo. 1983. Homage to Santa Anita: thermal sensitivity of sprint speed in agamid lizards. Evolution 37:1075–1084. Hertz, P. E., R. B. Huey, and R. D. Stevenson. 1993. Evaluating temperature regulation by ﬁeld-active ectotherms: the fallacy of the inappropriate question. American Naturalist 142:796–818. Hijmans, R. J., and C. H. Graham. 2006. The ability of climate envelope models to predict the effect of climate change on species distributions. Global Change Biology 12:1–10. Hilborn, R., and M. Mangel. 1997, The Ecological Detective. Princeton, Princeton University Press.

253

Hirshﬁeld, M. F., and D. W. Tinkle. 1975. Natural selection and the evolution of reproductive effort. Proceedings of the National Academy of Sciences 72:2227–2231. Hochochka, P. W., and G. N. Somero. 2002, Biochemical Adaptation. Oxford, Oxford University Press. Hoffmann, A., A. Anderson, and R. Hallas. 2002. Opposing clines for high and low temperature resistance in Drosophila melanogaster. Ecology Letters 5:614–618. Hoffmann, A. A. 1995. Acclimation: increasing survival at a cost. Trends in Ecology & Evolution 10:1–2. Hoffmann, A. A., J. G. Sorensen, and V. Loeschcke. 2003. Adaptation of Drosophila to temperature extremes: bringing together quantitative and molecular approaches. Journal of Thermal Biology 28:175–216. Hoffmann, A. A., and M. Watson. 1993. Geographical variation in the acclimation responses of Drosophila to temperature extremes. American Naturalist 142: S93-S113. Hoffmann, M. H., J. Tomiuk, H. Schmuths, C. Koch, and K. Bachmann. 2005. Phenological and morphological responses to different temperature treatments differ among a world-wide sample of accessions of Arabidopsis thaliana. Acta Oecologica 28:181–187. Holder, K. K., and J. J. Bull. 2001. Proﬁles of adaptation in two similar viruses. Genetics 159:1393–1404. Holland, K. N., R. W. Brill, R. K. C. Chang, J. R. Sibert, and D. A. Fournier. 1992. Physiological and behavioral thermoregulation in bigeye tuna (Thunnus obesus). Nature 358:410–412. Holt, R. D. 1990. The microevolutionary consequences of climate change. Trends in Ecology & Evolution 5: 311–315. Holt, R. D., M. Barﬁeld, and R. Gomulkiewicz. 2004. Temporal variation can facilitate niche evolution in harsh sink environments. American Naturalist 164:187–200. Holt, R. D., R. Gomulkiewicz, and M. Barﬁeld. 2003. The phenomenology of niche evolution via quantitative traits in a ‘black-hole’ sink. Proceedings of the Royal Society B 270:215–224. Hopkin, R. S., S. Qari, K. Bowler, D. Hyde, and M. Cuculescu. 2006. Seasonal thermal tolerance in marine Crustacea. Journal of Experimental Marine Biology and Ecology 331: 74–81. Howe, P. D., S. R. Bryant, and T. G. Shreeve. 2007. Predicting body temperature and activity of adult Polyommatus icarus using neural network models under current and projected climate scenarios. Oecologia 153:857–869. Howe, R. W. 1962. The effects of temperature and humidity on the oviposition rate of Tribolium castaneum (Hbst.)(Coleoptera, Tenebrionidae). Bulletin of Entomological Research 53:301–310.

254

REFERENCES

Howe, S. A., and A. T. Marshall. 2002. Temperature effects on calciﬁcation rate and skeletal deposition in the temperate coral, Plesiastrea versipora (Lamarck). Journal of Experimental Marine Biology and Ecology 275:63–81. Huey, R. B. 1974. Behavioral thermoregulation in lizards: importance of associated costs. Science 184:1001–1003. —. 1982. Temperature, physiology, and the ecology of reptiles, Pages 25–91 in C. Gans, and F. H. Pough, eds. Biology of the Reptilia. New York, Academic Press. —. 1987. Phylogeny, history, and the comparative method, Pages 76–98 in M. E. Feder, A. F. Bennett, W. W. Burggren, and R. B. Huey, eds. New Directions in Ecological Physiology. Cambridge, Cambridge University Press. —. 1991. Physiological consequences of habitat selection. American Naturalist 137:S91-S115. Huey, R. B., and A. F. Bennett. 1987. Phylogenetic studies of coadaptation: preferred temperatures versus optimal performance temperatures of lizards. Evolution 41: 1098–1115. Huey, R. B., and D. Berrigan. 1996. Testing evolutionary hypotheses of acclimation, Pages 205–237 in I. A. Johnston, and A. F. Bennett, eds. Animals and Temperature: Phenotypic and Evolutionary Adaptation. Cambridge, Cambridge University Press. —. 2001. Temperature, demography, and ectotherm ﬁtness. American Naturalist 158:204–210. Huey, R. B., D. Berrigan, G. W. Gilchrist, and J. C. Herron. 1999. Testing the adaptive signiﬁcance of acclimation: a strong inference approach. American Zoologist 39: 323–336. Huey, R. B., W. D. Crill, J. G. Kingsolver, and K. E. Weber. 1992. A method for rapid measurement of heat or cold resistance of small insects. Functional Ecology 6:489–494. Huey, R. B., G. W. Gilchrist, M. L. Carlson, D. Berrigan, and L. Serra. 2000. Rapid evolution of a geographic cline in size in an introduced ﬂy. Science 287:308–309. Huey, R. B., and P. E. Hertz. 1984. Is a jack-of-alltemperatures a master of none? Evolution 38:441–444. Huey, R. B., P. E. Hertz, and B. Sinervo. 2003. Behavioral drive versus behavioral inertia in evolution: a null model approach. American Naturalist 161:357–366. Huey, R. B., and J. G. Kingsolver. 1993. Evolution of resistance to high temperature in ectotherms. American Naturalist 142:S21-S46. Huey, R. B., P. H. Niewiarowski, J. Kaufmann, and J. C. Herron. 1989. Thermal biology of nocturnal ectotherms: is sprint performance of geckos maximal at low body temperatures? Physiological Zoology 62:488–504. Huey, R. B., L. Partridge, and K. Fowler. 1991. Thermal sensitivity of Drosophila melanogaster responds rapidly to laboratory natural selection. Evolution 45:751–756.

Huey, R. B., E. R. Pianka, and J. A. Hoffman. 1977. Seasonal variation in thermoregulatory behavior and body temperature of diurnal Kalahari lizards. Ecology 58:1066–1075. Huey, R. B., and M. Slatkin. 1976. Cost and beneﬁts of lizard thermoregulation. Quarterly Review of Biology 51:363–384. Huey, R. B., and R. D. Stevenson. 1979. Integrating thermal physiology and ecology of ectotherms: discussion of approaches. American Zoologist 19:357–366. Hughes, L. 2000. Biological consequences of global warming: is the signal already apparent? Trends in Ecology & Evolution 15:56–61. Hughes, N. F. 1998. A model of habitat selection by driftfeeding stream salmonids at different scales. Ecology 79:281–294. Hughes, N. F., and T. C. Grand. 2000. Physiological ecology meets the ideal-free distribution: predicting the distribution of size-structured ﬁsh populations across temperature gradients. Environmental Biology of Fishes 59:285–298. Hugie, D. M., and L. M. Dill. 1994. Fish and game: a gametheoretic approach to habitat selection by predators and prey. Journal of Fish Biology 45:151–169. Hulbert, A. J., and P. L. Else. 2000. Mechanisms underlying the cost of living in animals. Annual Review of Physiology 62:207–235. Humphries, M. M., D. L. Kramer, and D. W. Thomas. 2003b. The role of energy availability in mammalian hibernation: an experimental test in free-ranging eastern chipmunks. Physiological and Biochemical Zoology 76:180–186. Humphries, M. M., D. W. Thomas, and D. L. Kramer. 2003a. The role of energy availability in mammalian hibernation: a cost–beneﬁt approach. Physiological and Biochemical Zoology 76:165–179. Humphries, M. M., D. W. Thomas, and J. R. Speakman. 2002. Climate-mediated energetic constraints on the distribution of hibernating mammals. Nature 418:313–316. Hung, H. W., C. F. Lo, C. C. Tseng, and G. H. Kou. 1997. Antibody production in Japanese eels, Anguilla japonica Temminck & Schlegel. Journal of Fish Diseases 20: 195–200. Husak, J. F. 2006. Does speed help you survive? A test with collared lizards of different ages. Functional Ecology 20:174–179. Hutchinson, G. E. 1957. Concluding remarks. Cold Spring Harbor Symposia on Quantitative Biology 22:415–427. Hutchison, V. H., and J. D. Maness. 1979. Role of behavior in temperature acclimation and tolerance in ectotherms. American Zoologist 19:367–384.

REFERENCES

Imsland, A. K., T. M. Jonassen, S. O. Stefansson, S. Kadowaki, and M. H. G. Berntssen. 2000. Intraspeciﬁc differences in physiological efﬁciency of juvenile Atlantic halibut Hippoglossus hippoglossus L. Journal of the World Aquaculture Society 31:285–296. Intergovernmental Panel on Climate Change. 2000. Emissions Scenarios, A Special Report of the IPCC Working Group III. Intergovernmental Panel on Climate Change. 2007. Climate Change 2007: Synthesis Report, Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Isaac, L. A., and P. T. Gregory. 2004. Thermoregulatory behaviour of gravid and non-gravid female grass snakes (Natrix natrix) in a thermally limiting high-latitude environment. Journal of Zoology 264:403–409. Izem, R., and J. G. Kingsolver. 2005. Variation in continuous reaction norms: quantifying directions of biological interest. American Naturalist 166:277–289. Jackson, J. B. C. 1974. Biogeographic consequences of eurytopy and stenotopy among marine bivalves and their evolutionary signiﬁcance. American Naturalist 108: 541–560. Jaenicke, R. 1991. Protein stability and molecular adaptation to extreme conditions. European Journal of Biochemistry 202:715–728. —. 1993. Structure-function relationship of hyperthermophilic enzymes. ACS Symposium Series 516:53–67. James, A. C., R. B. R. Azevedo, and L. Partridge. 1995. Cellular basis and developmental timing in a size cline of Drosophila melanogaster. Genetics 140:659–666. —. 1997. Genetic and environmental responses to temperature of Drosophila melanogaster from a latitudinal cline. Genetics 146:881–890. James, A. C., and L. Partridge. 1995. Thermal evolution of rate of larval development in Drosophila melanogaster in laboratory and ﬁeld populations. Journal of Evolutionary Biology 8:315–330. Jansen, W., and R. H. Hesslein. 2004. Potential effects of climate warming on ﬁsh habitats in temperate zone lakes with special reference to Lake 239 of the experimental lakes area (ELA), north-western Ontario. Environmental Biology of Fishes 70:1–22. Jayne, B. C., and A. F. Bennett. 1990. Selection on locomotor performance capacity in a natural population of garter snakes. Evolution 44:1204–1229. Jeffery, W. R. 2006. Regressive evolution of pigmentation in the caveﬁsh Astyanax. Israel Journal of Ecology & Evolution 52:405–422. Ji, X., W. G. Du, and P. Y. Sun. 1996. Body temperature, thermal tolerance and inﬂuence of temperature on sprint speed and food assimilation in adult grass lizards,

255

Takydromus septentrionalis. Journal of Thermal Biology 21:155–161. Ji, X., C.-X. Lin, L.-H. Lin, Q.-B. Qiu, and Y. Du. 2007. Evolution of vivparity in warm-climate lizards: an experimental test of the maternal manipulation hypothesis Journal of Evolutionary Biology 20: 1037–1045. Ji, X., X. Zheng, Y. Xu, and R. Sun. 1995. Some aspects of thermal biology of the skink (Eumeces chinensis). Acta Zoologica Sinica 41:268–274. Jiang, L., and A. Kulczycki. 2004. Competition, predation and species responses to environmental change. Oikos 106:217–224. Jiang, L., and P. J. Morin. 2004. Temperature-dependent interactions explain unexpected responses to environmental warming in communities of competitors. Journal of Animal Ecology 73:569–576. Jin, Y. M., and A. L. DeVries. 2006. Antifreeze glycoprotein levels in Antarctic notothenioid ﬁshes inhabiting different thermal environments and the effect of warm acclimation. Comparative Biochemistry and Physiology B 144:290–300. Joern, A., B. J. Danner, J. D. Logan, and W. Wolesensky. 2006. Natural history of mass-action in predator-prey models: a case study from wolf spiders and grasshoppers. American Midland Naturalist 156:52–64. Johansson, J. 2008. Evolutionary responses to environmental changes: how does competition affect adaptation? Evolution 62:421–435. John-Alder, H. B., and A. F. Bennett. 1987. Thermal adaptations in lizard muscle function. Journal of Comparative Physiology B 157:241–252. John-Alder, H. B., P. J. Morin, and S. Lawler. 1988. Thermal physiology, phenology, and distribution of tree frogs. American Naturalist 132:506–520. Johnsen, O., C. G. Fossdal, N. Nagy, J. Molmann, O. G. Daehlen, and T. Skroppa. 2005. Climatic adaptation in Picea abies progenies is affected by the temperature during zygotic embryogenesis and seed maturation. Plant, Cell and Environment 28:1090–1102. Johnson, J. B., and K. S. Omland. 2004. Model selection in ecology and evolution. Trends in Ecology & Evolution 19:101–108. Johnson, T. P., and A. F. Bennett. 1995. The thermal acclimation of burst escape performance in ﬁsh: an integrated study of molecular and cellular physiology and organismal performance. Journal of Experimental Biology 198:2165–2175. Johnston, I. A., and R. S. Wilson. 2006. Temperatureinduced developmental plasticity in ectotherms, Pages 124–138 in S. J. Warburton, W. W. Burggren, B. Pelster, C. L. Reiber, and J. Spicer, eds. Comparative Developmental Physiology.

256

REFERENCES

Jonassen, T. M., A. K. Imsland, R. Fitzgerald, S. W. Bonga, E. V. Ham, G. Naevdal, M. O. Stefansson et al. 2000. Geographic variation in growth and food conversion efﬁciency of juvenile Atlantic halibut related to latitude. Journal of Fish Biology 56:279–294. Jones, A. G., S. J. Arnold, and R. Borger. 2003. Stability of the G-matrix in a population experiencing pleiotropic mutation, stabilizing selection, and genetic drift. Evolution 57:1747–1760. Jones, A. G., S. J. Arnold, and R. Burger. 2004. Evolution and stability of the G-matrix on a landscape with a moving optimum. Evolution 58:1639–1654. Jonsson, B., T. Forseth, A. J. Jensen, and T. F. Naesje. 2001. Thermal performance of juvenile Atlantic salmon, Salmo salar L. Functional Ecology 15:701–711. Jump, A. S., and J. Penuelas. 2005. Running to stand still: adaptation and the response of plants to rapid climate change. Ecology Letters 8:1010–1020. Kalberer, S. R., M. Wisniewski, and R. Arora. 2006. Deacclimation and reacclimation of cold-hardy plants: current understanding and emerging concepts. Plant Science 171:3–16. Kamil, A. C., J. R. Krebs, and H. R. Pulliam. 1987. Foraging Behavior, Page 676. New York, Plenum Press. Kammer, A. E. 1981. Physiological mechanisms of thermoregulation, Pages 115–158 in B. Heinrich, ed. Insect Thermoregulation. New York, Wiley. Karl, I., and K. Fischer. 2008. Why get big in the cold? Towards a solution to a life-history puzzle. Oecologia 155:215–225. Karl, I., S. A. Janowitz, and K. Fischer. 2008. Altitudinal life-history variation and thermal adaptation in the copper butterﬂy Lycaena tityrus. Oikos 117:778–788. Karlsson, B., and H. Van Dyck. 2005. Does habitat fragmentation affect temperature-related life-history traits? A laboratory test with a woodland butterﬂy. Proceedings of the Royal Society of London B 272:1257–1263. Karlsson, B., and C. Wiklund. 2005. Butterﬂy life history and temperature adaptations: dry open habitats select for increased fecundity and longevity. Journal of Animal Ecology 74:99–104. Kassen, R. 2002. The experimental evolution of specialists, generalists, and the maintenance of diversity. Journal of Evolutionary Biology 15:173–190. Kauffman, A. S., M. J. Paul, and I. Zucker. 2004. Increased heat loss affects hibernation in golden-mantled ground squirrels. American Journal of Physiology 287: R167– R173. Kaufmann, J. S., and A. F. Bennett. 1989. The effect of temperature and thermal acclimation on locomotor performance in Xantusia vigilis, the desert night lizard. Physiological Zoology 62:1047–1058.

Kaufmann, R., and F. H. Pough. 1982. The effect of temperature upon the efﬁciency of assimilation of preformed water by the desert iguana (Dipsosaurus dorsalis). Comparative Biochemistry and Physiology 72A: 221–224. Kawecki, T. J., and D. Ebert. 2004. Conceptual issues in local adaptation. Ecology Letters 7:1225–1241. Kearney, M., and W. P. Porter. 2004. Mapping the fundamental niche: physiology, climate, and the distribution of a nocturnal lizard. Ecology 85:3119–3131. Kelty, J. D., and R. E. Lee. 2001. Rapid cold-hardening of Drosophila melanogaster (Diptera: Drosophilidae) during ecologically based thermoperiodic cycles. Journal of Experimental Biology 204:1659–1666. Kemp, D. J., and A. K. Krockenberger. 2002. A novel method of behavioural thermoregulation in butterﬂies. Journal of Evolutionary Biology 15:922–929. —. 2004. Behavioural thermoregulation in butterﬂies: the interacting effects of body size and basking posture in Hypolimnas bolina (L.) (Lepidoptera: Nymphalidae). Australian Journal of Zoology 52:229–236. Kemp, T. S. 2006. The origin of mammalian endothermy: a paradigm for the evolution of complex biological structure. Zoological Journal of the Linnean Society 147:473–488. Kerr, G. D., and C. M. Bull. 2006. Interactions between climate, host refuge use, and tick population dynamics. Parasitology Research 99:214–222. Kerr, J. T., and M. Ostrovsky. 2003. From space to species: ecological applications for remote sensing. Trends in Ecology & Evolution 18:299–305. Kessler, K., and W. Lampert. 2004. Fitness optimization of Daphnia in a trade-off between food and temperature. Oecologia 140:381–387. Ketterson, E. D., and V. Nolan. 1992. Hormones and life histories: an integrative approach. American Naturalist 140:S33-S62. Kiefer, M. C., M. Van Sluys, and C. F. D. Rocha. 2005. Body temperatures of Tropidurus torquatus (Squamata, Tropiduridae) from coastal populations: do body temperatures vary along their geographic range? Journal of Thermal Biology 30:449–456. Kimura, M. T. 2004. Cold and heat tolerance of drosophilid ﬂies with reference to their latitudinal distributions. Oecologia 140:442–449. Kindlmann, P., A. F. G. Dixon, and I. Dostalkova. 2001. Role of ageing and temperature in shaping reaction norms and fecundity functions in insects. Journal of Evolutionary Biology 14:835–840. Kingsolver, J. G. 1983. Thermoregulation and ﬂight in Colias butterﬂies: elevational patterns and mechanistic limitations. Ecology 64:534–545.

REFERENCES

—. 1987. Evolution and coadaptation of thermoregulatory behavior and wing pigmentation pattern in pierid butterﬂies. Evolution 41:472–490. —. 2000. Feeding, growth, and the thermal environment of cabbage white caterpillars, Pieris rapae L. Physiological and Biochemical Zoology 73:621–628. —. 2001. Mechanisms and patterns of selection on performance curves: thermal sensitivity of caterpillar growth in D. Atkinson, and M. Thorndyke, eds. Environment and Animal Development: Genes, Life Histories and Plasticity Oxford, BIOS Scientiﬁc Publishers, Ltd. —. 2007. Relating environmental variation to selection on reaction norms: an experimental test American Naturalist 169:163–174. Kingsolver, J. G., and R. Gomulkiewicz. 2003. Environmental variation and selection on performance curves. Integrative and Comparative Biology 43:470–477. Kingsolver, J. G., R. Gomulkiewicz, and P. A. Carter. 2001a. Variation, selection and evolution of function-valued traits. Genetica 112:87–104. Kingsolver, J. G., H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S. N. Vignieri, C. E. Hill, A. Hoang et al. 2001b. The strength of phenotypic selection in natural populations. American Naturalist 157:245–261. Kingsolver, J. G., and R. B. Huey. 1998. Evolutionary analyses of morphological and physiological plasticity in thermally variable environments. American Zoologist 38:545–560. Kingsolver, J. G., K. R. Massie, G. J. Ragland, and M. H. Smith. 2007. Rapid population divergence in thermal reaction norms for an invading species: breaking the temperature-size rule. Journal of Evolutionary Biology 20:892–900. Kingsolver, J. G., G. J. Ragland, and J. G. Shlichta. 2004. Quantitative genetics of continuous reaction norms: thermal sensitivity of caterpillar growth rates. Evolution 58:1521–1529. Kingsolver, J. G., J. G. Shlichta, G. J. Ragland, and K. R. Massie. 2006. Thermal reaction norms for caterpillar growth depend on diet. Evolutionary Ecology Research 8:703–715. Kingsolver, J. G., and H. A. Woods. 1997. Thermal sensitivity of growth and feeding in Manduca sexta caterpillars. Physiological Zoology 70:631–638. —. 1998. Interactions of temperature and dietary protein concentration in growth and feeding of Manduca sexta caterpillars. Physiological Entomology 23:354–359. Kirkpatrick, M., and N. H. Barton. 1997. Evolution of a species’ range. American Naturalist 150:1–23. Kirschbaum, M. U. F. 2004. Direct and indirect climate change effects on photosynthesis and transpiration. Plant Biology 6:242–253.

257

Klok, C. J., and S. L. Chown. 2005. Inertia in physiolgical traits: Embryonopsis halticella caterpillars (Yponomeutidae) across the Antarctic polar frontal zone. Journal of Insect Physiology 51:87–97. Klok, C. J., B. J. Sinclair, and S. L. Chown. 2004. Upper thermal tolerance and oxygen limitation in terrestrial arthropods. Journal of Experimental Biology 207: 2361–2370. Kluger, M. J. 1979. Fever in ectotherms: evolutionary implications. American Zoologist 19:295–304. Kluger, M. J., W. Kozak, C. A. Conn, L. R. Leon, and D. Soszynski. 1998. Role of fever in disease. Annals of the New York Academy of Sciences 856:224–233. Knies, J. L., R. Izem, K. L. Supler, J. G. Kingsolver, and C. L. Burch. 2006. The genetic basis of thermal reaction norm evolution in lab and natural phage populations. PLoS Biology 4:1–8. Kopp, M., and S. Gavrilets. 2006. Multilocus genetics and the coevolution of quantitative traits. Evolution 60: 1321–1336. Kopp, M., and J. Hermisson. 2006. The evolution of genetic architecture under frequency-dependent disruptive selection. Evolution 60:1537–1550. Koteja, P. 2000. Energy assimilation, parental care and the evolution of endothermy. Proceedings of the Royal Society B 267:479–484. —. 2004. The evolution of concepts on the evolution of endothermy in birds and mammals. Physiological and Biochemical Zoology 77:1043–1050. Kozłowski, J. 1992. Optimal allocation of resources to growth and reproduction: implications for age and size at maturity. Trends in Ecology & Evolution 7:15–19. Kozłowski, J., M. Czarnołeski, and M. Dañko. 2004. Can optimal resource allocation models explain why ectotherms grow larger in cold? Integrative and Comparative Biology 44:480–493. Krebs, R. A., and M. E. Feder. 1997. Natural variation in the expression of the heat-shock protein Hsp70 in a population of Drosophila melanogaster and its correlation with tolerance of ecologically relevant thermal stress. Evolution 51:173–179. —. 1998. Experimental manipulation of the cost of thermal acclimation in Drosophila melanogaster. Biological Journal of the Linnean Society 63:593–601. Krebs, R. A., and V. Loeschcke. 1994a. Costs and beneﬁts of activation of the heat-shock response in Drosophila melanogaster. Functional Ecology 8:730–737. —. 1994b. Effects of exposure to short-term heat stress on ﬁtness components in Drosophila melanogaster. Journal of Evolutionary Biology 7:39–49. —. 1995. Resistance to thermal stress in adult Drosophila buzzatii: acclimation and variation among populations.

258

REFERENCES

Biological Journal of the Linnean Society 56: 505–515. —. 1996. Acclimation and selection for increased resistance to thermal stress in Drosophila buzzatii. Genetics 142: 471–479. —. 1999. A genetic analysis of the relationship between lifehistory variation and heat-shock tolerance in Drosophila buzzatii. Heredity 83:46–53. Kreuger, B., and D. A. Potter. 2001. Diel feeding activity and thermoregulation by Japanese beetles (Coleoptera: Scarabaeidae) within host plant canopies. Environmental Entomology 30:172–180. Krochmal, A. R., and G. S. Bakken. 2003. Thermoregulation is the pits: use of thermal radiation for retreat site selection by rattlesnakes. Journal of Experimental Biology 206:2539–2545. Kronfeld-Schor, N., and T. Dayan. 2003. Partitioning of time as an ecological resource. Annual Review of Ecology Evolution and Systematics 34:153–181. Ksiazek, A., M. Konarzewski, and I. B. Lapo. 2004. Anatomic and energetic correlates of divergent selection for basal metabolic rate in laboratory mice. Physiological and Biochemical Zoology 77:890–899. Kudo, I., M. Miyamoto, Y. Noiri, and Y. Maita. 2000. Combined effects of temperature and iron on the growth and physiology of the marine diatom Phaeodactylum tricornutum (Bacillariophyceae). Journal of Phycology 36:1096–1102. Lacey, E. P., and D. Herr. 2005. Phenotypic plasticity, parental effects and parental care in plants? I. An examination of spike reﬂectance in Plantago lanceolata (Plantaginaceae). American Journal of Botany 92:920–930. Ladyman, M., and D. Bradshaw. 2003. The inﬂuence of dehydration on the thermal preferences of the western tiger snake, Notechis scutatus. Journal of Comparative Physiology B 173:239–246. Lagerspetz, K. Y. H. 2006. What is thermal acclimation? Journal of Thermal Biology 31:332–336. Lagerspetz, K. Y. H., and L. A. Vainio. 2006. Thermal behaviour of crustaceans. Biological Reviews 81: 237–258. Lakoff, G., and R. E. Núñez. 2000, Where Mathematics Comes From: How the Embodied Mind Brings Mathematics into Being. New York, Basic Books. Lamb, R. J., and G. H. Gerber. 1985. Effects of temperature on the development, growth, and survival of larvae and pupae of a north-temperate chrysomelid beetle. Oecologia 67:8–18. Lamb, R. J., P. A. Mackay, and G. H. Gerber. 1987. Are development and growth of pea aphids, Acyrthosiphon pisum, in North America adapted to local temperatures? Oecologia 72:170–177.

Lampert, W. 1989. The adaptive signiﬁcance of diel vertical migration of zooplankton. Functional Ecology 3: 21–27. —. 2005. Vertical distribution of zooplankton: density dependence and evidence for an ideal free distribution with costs. BMC Biology 3. Lampert, W., E. McCauley, and B. F. J. Manly. 2003. Tradeoffs in the vertical distribution of zooplankton: ideal free distribution with costs? Proceedings of the Royal Society B 270:765–773. Lande, R. 1979. Quantitative genetic analysis of multivariate evolution, applied to brain-body size allometry. Evolution 33:402–416. —. 1984. The genetic correlation between characters maintained by selection, linkage and inbreeding. Genetical Research 44:309–320. Lande, R., and S. J. Arnold. 1983. The measurement of selection on correlated characters. Evolution 37:1210–1226. Lankford, T. E., J. M. Billerbeck, and D. O. Conover. 2001. Evolution of intrinsic growth and energy acquisition rates. II. Trade-offs with vulnerability to predation in Menidia menidia. Evolution 55:1873–1881. Lankford, T. E., and T. E. Targett. 2001. Physiological performance of young-of-the-year Atlantic croakers from different Atlantic coast estuaries: implications for stock structure. Transactions of the American Fisheries Society 130:367–375. Laptikhovsky, V. 2006. Latitudinal and bathymetric trends in egg size variation: a new look at Thorson’s and Rass’s rules. Marine Ecology 27:7–14. Larimore, R. W., and M. J. Duever. 1968. Effects of temperature acclimation on swimming ability of smallmouth bass fry. Transactions of the American Fisheries Society 97:175–184. Larsson, S., T. Forseth, I. Berglund, A. J. Jensen, I. Naslund, J. M. Elliott, and B. Jonsson. 2005. Thermal adaptation of Arctic charr: experimental studies of growth in eleven charr populations from Sweden, Norway and Britain. Freshwater Biology 50:353–368. Laugen, A. T., A. Laurila, K. I. Jonsson, F. Soderman, and J. Merila. 2005. Do common frogs (Rana temporaria) follow Bergmann’s rule? Evolutionary Ecology Research 7: 717–731. Laugen, A. T., A. Laurila, and J. Merila. 2003b. Latitudinal and temperature-dependent variation in embryonic development and growth in Rana temporaria. Oecologia 135:548–554. Laugen, A. T., A. Laurila, K. Rasanen, and J. Merila. 2003a. Latitudinal countergradient variation in the common frog (Rana temporaria) development rates - evidence for local adaptation. Journal of Evolutionary Biology 16:996–1005.

REFERENCES

Law, R. 1979. Optimal life histories under age-speciﬁc predation. American Naturalist 114:399–417. Le Galliard, J. F., M. Le Bris, and J. Clobert. 2003. Timing of locomotor impairment and shift in thermal preferences during gravidity in a viviparous lizard. Functional Ecology 17:877–885. Lebedeva, L. I., and T. N. Gerasimova. 1985. Peculiarities of Philodina roseola (Ehrbg.) (Rotatoria, Bdelloida): growth and reproduction under various temperature conditions. Internationale Revue der gesamten Hydrobiologie 70:509–525. Lee, R. E., and J. P. Costanzo. 1998. Biological ice nucleation and ice distribution in cold-hardy ectothermic animals. Annual Review of Physiology 60:55–72. Lee, R. E., K. Damodaran, S. X. Yi, and G. A. Lorigan. 2006. Rapid cold-hardening increases membrane ﬂuidity and cold tolerance of insect cells. Cryobiology 52:459–463. Lefﬂer, C. W. 1972. Some effects of temperature on the growth and metabolic rate of juvenile blue crabs, Callinectes sapidus, in the laboratory. Marine Biology 14: 104–110. Lek, S., M. Delacoste, P. Baran, I. Dimopoulos, J. Lauga, and S. Aulagnier. 1996. Application of neural networks to modelling nonlinear relationships in ecology. Ecological Modelling 90:39–52. Lek, S., and J. F. Guégan. 1999. Artiﬁcial neural networks as a tool in ecological modelling, an introduction. Ecological Modelling 120:65–73. —. 2000, Artiﬁcial Neuronal Networks: Application to Ecology and Evolution. Berlin, Springer. Lemos-Espinal, J. A., G. R. Smith, and R. E. Ballinger. 1997. Thermal ecology of the lizard, Sceloporus gadoviae, in an arid tropical scrub forest. Journal of Arid Environments 35:311–319. Lenormand, T. 2002. Gene ﬂow and the limits to natural selection. Trends in Ecology & Evolution 17:183–189. Leroi, A. M., A. F. Bennett, and R. E. Lenski. 1994. Temperature-acclimation and competitive ﬁtness: an experimental test of the beneﬁcial acclimation assumption. Proceedings of the National Academy of Sciences 91:1917–1921. Levins, R. 1968, Evolution in Changing Environments: Some Theoretical Explorations. Princeton, Princeton University Press. Li, D. Q. 2002. The combined effects of temperature and diet on development and survival of a crab spider, Misumenops tricuspidatus (Fabricius) (Araneae: Thomisidae). Journal of Thermal Biology 27:83–93. Li, X. C., and L. Z. Wang. 2005. Effect of temperature and thermal acclimation on locomotor performance of Macrobiotus harmsworthi Murray (Tardigrada, Macrobiotidae). Journal of Thermal Biology 30:588–594.

259

Licht, P. 1967. Thermal adaptation in the enzymes of lizards in relation to preferred body temperatures, Pages 131–146 in C. L. Prosser, ed. Molecular Mechanisms of Temperature Adaptation. Washington, D. C., American Association for the Advancement of Science. —. 1968. Response of the thermal preferendum and heat resistance to thermal acclimation under different photoperiods in the lizard Anolis carolinensis. American Midland Naturalist 79:149–158. Licht, P., W. R. Dawson, and V. H. Shoemaker. 1966a. Heat resistance of some Australian lizards. Copeia 1966: 162–169. Licht, P., W. R. Dawson, V. H. Shoemaker, and A. R. Main. 1966b. Observations on the thermal relations of western Australian lizards. Copeia 1966:97–110. Lillywhite, H. B., and C. A. Navas. 2006. Animals, energy, and water in extreme environments: perspectives from Ithala 2004. Physiological and Biochemical Zoology 79:265–273. Lima, F. P., P. A. Ribeiro, N. Queiroz, S. J. Hawkins, and A. M. Santos. 2007a. Do distributional shifts in northern and southern species of algae match the warming pattern? Global Change Biology 13:2592–2604. Lima, F. P., P. A. Ribeiro, N. Queiroz, R. Xavier, P. Tarroso, S. J. Hawkins, and A. M. Santos. 2007b. Modelling past and present geographical distribution of the marine gastropod Patella rustica as a tool for exploring responses to environmental change. Global Change Biology 13:2065– 2077. Lima, S. L. 2002. Putting predators back into behavioral predator-prey interactions. Trends in Ecology & Evolution 17:70–75. Lima, S. L., and L. M. Dill. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology 68:619–640. Lindgren, B., and A. Laurila. 2005. Proximate causes of adaptive growth rates: growth efﬁciency variation among latitudinal populations of Rana temporaria. Journal of Evolutionary Biology 18:820–828. Loeschcke, V., J. Bundgaard, and J. S. F. Barker. 1999. Reaction norms across and genetic parameters at different temperatures for thorax and wing size traits in Drosophila aldrichi and D. buzzatii. Journal of Evolutionary Biology 12:605–623. Loeschcke, V., and R. A. Krebs. 1996. Selection for heatshock resistance in larval and in adult Drosophila buzzatii: comparing direct and indirect responses. Evolution 50:2354–2359. Loeschcke, V., R. A. Krebs, and J. S. F. Barker. 1994. Genetic variation for resistance and acclimation to high temperature stress in Drosophila buzzatii. Biological Journal of the Linnean Society 52:83–92.

260

REFERENCES

Loeschcke, V., and J. G. Sorenson. 2005. Acclimation, heat shock and hardening—a response from evolutionary biology. Journal of Thermal Biology 30:255–257. Lonsdale, D. J., and J. S. Levinton. 1985. Latitudinal differentiation in copepod growth: an adaptation to temperature. Ecology 66:1397–1407. —. 1989. Energy budgets of latitudinally separated Scottolana canadensis (Copepoda, Harpacticoida). Limnology and Oceanography 34:324–331. Loose, C. J., and P. Dawidowicz. 1994. Trade-offs in diel vertical migration by zooplankton: the costs of predator avoidance. Ecology 75:2255–2263. Lorenzon, P., J. Clobert, A. Oppliger, and H. John-Alder. 1999. Effect of water constraint on growth rate, activity and body temperature of yearling common lizard (Lacerta vivipara). Oecologia 118:423–430. Lourdais, O., B. Heulin, and D. F. Denardo. 2008. Thermoregulation during gravidity in the children’s python (Antaresia childreni): a test of the preadaptation hypothesis for maternal thermophily in snakes. Biological Journal of the Linnean Society 93:499–508. Luo, L. G., Y. F. Qu, and X. Ji. 2006. Thermal dependence of food assimilation and sprint speed in a lacertid lizard Eremias argus from northern China. Acta Zoologica 52:256–262. Lustick, S. I. 1983. Cost–beneﬁt of thermoregulation in birds: inﬂuences of posture, microhabitat selection, and color, Pages 265–294 in W. P. Aspey, and S. I. Lustick, eds. Behavioral Energetics. Columbus, Ohio State University Press. Lutterschmidt, W. I., and V. H. Hutchison. 1997a. The critical thermal maximum: data to support the onset of spasms as the deﬁnitive end point. Canadian Journal of Zoology 75:1553–1560. —. 1997b. The critical thermal maximum: history and critique. Canadian Journal of Zoology 75:1561–1574. Lutterschmidt, W. I., and H. K. Reinert. 1990. The effect of ingested transmitters upon the temperature preference of the northern water snake, Nerodia s. sipedon. Herpetologica 46:39–42. Lynch, M., and W. Gabriel. 1987. Environmental tolerance. American Naturalist 129:283–303. Lynch, M., and R. Lande. 1993. Evolution and extinction in response to environmental change, Pages 234–250 in P. Kareiva, J. Kingsolver, and R. Huey, eds. Biotic Interactions and Global Change. Sunderland, Sinauer Assocs., Inc. Lynch, M., and B. Walsh. 1998, Genetics and Analysis of Quantitative Traits. Sunderland, Sinauer Associates, Inc. Lysenko, S., and J. E. Gillis. 1980. The effect of ingestive status on the thermoregulatory behavior of Thamnophis

sirtalis sirtalis and Thamnophis sirtalis parietalis. Journal of Herpetology 14:155–159. Mac, M. J. 1985. Effects of ration size on preferred temperature of lake charr Salvelinus namaycush. Environmental Biology of Fishes 14:227–231. Macario, A. J. L., and E. C. de Macario. 2007. Molecular chaperones: multiple functions, pathologies, and potential applications. Frontiers in Bioscience 12:2588–2600. Magnum, C. P., and P. W. Hochachaka. 1998. New directions in comparative physiology and biochemistry: mechanisms, adaptations, and evolution. Physiological Zoology 71:471–484. Magnuson, J. J., L. B. Crowder, and P. A. Medvick. 1979. Temperature as an ecological resource. American Zoologist 19:331–343. Mangel, M., and C. W. Clark. 1988, Dynamic Modeling in Behavioral Ecology. Princeton, Princeton University Press. Martin, J., and P. Lopez. 2001. Repeated predatory attacks and multiple decisions to come out from a refuge in an alpine lizard. Behavioral Ecology 12:386–389. Martin, T. L., and R. B. Huey. 2008. Why “suboptimal” is optimal: Jensen’s inequality and ectotherm thermal preferences. American Naturalist 171:E102–E118. Martins, E. P. 2000. Adaptation and the comparative method. Trends in Ecology & Evolution 15:296–299. Martins, E. P., and T. Garland. 1991. Phylogenetic analyses of the correlated evolution of continuous characters: a simulation study. Evolution 45:534–557. Martins, E. P., and T. F. Hansen. 1996. The statistical analysis of interspeciﬁc data: a review and evaluation of phylogenetic comparative methods, Pages 22–75 in E. P. Martins, ed. Phylogenies and the Comparative Method in Animal Behavior. Oxford, Oxford University Press. —. 1997. Phylogenies and the comparative method: a general approach to incorporating phylogenetic information into the analysis of interspeciﬁc data. American Naturalist 149:646–667. Marx, J. C., T. Collins, S. D’Amico, G. Feller, and C. Gerday. 2007. Cold-adapted enzymes from marine Antarctic microorganisms. Marine Biotechnology 9:293–304. Masaki, S. 1967. Geographic variation and climatic adaptation in a ﬁeld cricket (Orthoptera: Gryllidae). Evolution 21:725–741. Mathies, T., and R. M. Andrews. 1997. Inﬂuence of pregnancy on the thermal biology of the lizard, Sceloporus jarrovi: why do pregnant females exhibit low body temperatures? Functional Ecology 11:498–507. May, M. L. 1995. Simultaneous control of head and thoracic temperature by the green darner dragonﬂy Anax junius

REFERENCES

(Odonata, Aeshnidae). Journal of Experimental Biology 198:2373–2384. Maynard Smith, J. 1972. Game theory and the evolution of ﬁghting, Pages 8–28 in J. Maynard Smith, ed. On Evolution. Edinburgh, Edinburgh University Press. —. 1976. Evolution and the theory of games. American Scientist 64:41–45. —. 1978. Optimization theory in evolution. Annual Review of Ecology and Systematics 9:31–56. —. 1982, Evolution and the Theory of Games. Cambridge, Cambridge University Press. Maynard Smith, J., and G. R. Price. 1973. The logic of animal conﬂict. Nature 246:15–18. McCabe, J., and L. Partridge. 1997. An interaction between environmental temperature and genetic variation for body size for the ﬁtness of adult female Drosophila melanogaster. Evolution 51:1164–1174. McConnell, E., and A. G. Richards. 1955. How fast can a cockroach run? Bulletin of the Brooklyn Entomological Society 50:36–43. McCullough, E. C., and W. P. Porter. 1971. Computing clear day solar radiation spectra for terrestrial ecological environment. Ecology 52:1008–1015. McGinley, M. A., D. H. Temme, and M. A. Geber. 1987. Parental investment in offspring in variable environments: theoretical and empirical considerations. American Naturalist 130:370–398. McKechnie, A. E., G. Kortner, and B. G. Lovegrove. 2006. Thermoregulation under semi-natural conditions in speckled mousebirds: the role of communal roosting. African Zoology 41:155–163. McKechnie, A. E., and B. G. Lovegrove. 2002. Avian facultative hypothermic responses: a review. Condor 104:705–724. McLean, M. A., M. J. Angilletta, and K. S. Williams. 2005. If you can’t stand the heat, stay out of the city: thermal reaction norms of chitinolytic fungi in an urban heat island. Journal of Thermal Biology 30:384–391. McNab, B. K. 2002, Physiological Ecology of Vertebrates. Ithaca, Cornell University Press. McPeek, M. A. 2004. The growth/predation risk tradeoff: so what is the mechanism? American Naturalist 163: E88–E111. Measey, G. J., and S. Van Dongen. 2006. Bergmann’s rule and the terrestrial caecilian Schistometopum thomense (Amphibia: Gymnophiona: Caeciiiidae). Evolutionary Ecology Research 8:1049–1059. Medvick, P. A., J. J. Magnuson, and S. Sharr. 1981. Behavioral thermoregulation and social interactions of bluegills, Lepomis macrochirus. Copeia:9–13.

261

Meffe, G. K. 1990. Offspring size variation in eastern mosquitoﬁsh (Gambusia holbrooki: Poeciliidae) from contrasting thermal environments. Copeia 1990:10–18. Meiri, S., and T. Dayan. 2003. On the validity of Bergmann’s rule. Journal of Biogeography 30:331–351. Mejia-Ortiz, L. M., and R. G. Hartnoll. 2005. Modiﬁcations of eye structure and integumental pigment in two cave crayﬁsh. Journal of Crustacean Biology 25:480–487. Menéndez, R., A. González-Megías, O. T. Lewis, M. R. Shaw, and C. D. Thomas. 2008. Escape from natural enemies during climate-driven range expansion: a case study. Ecological Entomology 33:413–421. Merila, J., A. Laurila, A. T. Laugen, K. Rasanen, and M. Pahkala. 2000. Plasticity in age and size at metamorphosis in Rana temporaria – comparison of high and low latitude populations. Ecography 23:457–465. Merrill, R. M., D. Gutiérrez, O. T. Lewis, J. Gutiérrez, S. B. Díez, and R. J. Wilson. 2008. Combined effects of climate and biotic interactions on the elevational range of a phytophagous insect. Journal of Animal Ecology 77:145–155. Michod, R. E. 1979. Evolution of life histories in response to age-speciﬁc mortality factors. American Naturalist 113:531–550. Miles, D. B., and A. E. Dunham. 1993. Historical perspectives in ecology and evolutionary biology: the use of phylogenetic comparative analyses. Annual Review of Ecology and Systematics 24:587–619. Miller, K., and G. C. Packard. 1977. An altitudinal cline in critical thermal maxima of chorus frogs (Pseudacris triseriata). American Naturalist 111:267–277. Miller, S. R., and R. W. Castenholz. 2000. Evolution of thermotolerance in hot spring cyanobacteria of the genus Synechococcus. Applied and Environmental Microbiology 66:4222–4229. Minoli, S. A., and C. R. Lazzari. 2003. Chronobiological basis of thermopreference in the haematophagous bug Triatoma infestans. Journal of Insect Physiology 49: 927–932. Misra, R. K., and E. C. R. Reeve. 1964. Clines in body dimensions in populations of Drosophila subobscura. Genetical Research 5:240–256. Mitchell, J. W. 1976. Heat transfer from spheres and other animal forms. Biophysical Journal 16:561–569. Mitchell, M., and C. E. Taylor. 1999. Evolutionary computation: an overview. Annual Review of Ecology and Systematics 30:593–616. Mitchell, S. E., and W. Lampert. 2000. Temperature adaptation in a geographically widespread zooplankter, Daphnia magna. Journal of Evolutionary Biology 13:371–382.

262

REFERENCES

Mitchell, W. A., and T. J. Valone. 1990. The optimization research program: studying adaptations by their function. Quarterly Review of Biology 65:43–52. Mitten, J. B. 1997, Selection in Natural Populations. Oxford, Oxford University Press. Møller, H., R. H. Smith, and R. M. Sibly. 1989. Evolutionary demography of a bruchid beetle. II. Physiological manipulations. Functional Ecology 3:683–691. Mondal, S., and U. Rai. 2001. In vitro effect of temperature on phagocytic and cytotoxic activities of splenic phagocytes of the wall lizard, Hemidactylus ﬂaviviridis. Comparative Biochemistry and Physiology A 129: 391–398. Mongold, J. A., A. F. Bennett, and R. E. Lenski. 1999. Evolutionary adaptation to temperature. VII. Extension of the upper thermal limit of Escherichia coli. Evolution 53:386–394. Monteith, J. J., and M. H. Unsworth. 1990, Principles of Environmental Physics. London, Edward Arnold. Mooney, H. A. 1980. Photosynthetic plasticity of populations of Heliotropium curassavicum L. originating from differing thermal regimes. Oecologia 45:372–376. Moore, C. M., and L. M. Sievert. 2001. Temperaturemediated characteristics of the dusky salamander (Desmognathus fuscus) of southern Appalachia. Journal of Thermal Biology 26:547–554. Moore, J. A. 1940. Stenothermy and eurythermy of animals in relation to habitat. American Naturalist 74:188–192. Moran, N. A. 1992. The evolutionary maintenance of alternative phenotypes. American Naturalist 139: 971–989. Morgan, I. J., and N. B. Metcalfe. 2001. The inﬂuence of energetic requirements on the preferred temperature of overwintering juvenile Atlantic salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences 58:762–768. Morrison, C., and J.-M. Hero. 2003b. Geographic variation in life-history characteristics of amphibians: a review. Journal of Animal Ecology 72:270–279. Morrison, C., and J. M. Hero. 2003a. Altitudinal variation in growth and development rates of tadpoles of Litoria chloris and Litoria pearsoniana in southeast Queensland, Australia. Journal of Herpetology 37:59–64. Mousseau, T. A. 1997. Ectotherms follow the converse to Bergmann’s Rule. Evolution 51:630–632. Mullens, D. P., and V. H. Hutchison. 1992. Diel, seasonal, postprandial and food-deprived thermoregulatory behavior in tropical toads (Bufo marinus). Journal of Thermal Biology 17:63–67. Munoz, J. L. P., G. R. Finke, P. A. Camus, and F. Bozinovic. 2005. Thermoregulatory behavior, heat gain and thermal tolerance in the periwinkle Echinolittorina peruviana

in central Chile. Comparative Biochemistry and Physiology A 142:92–98. Munro, D., D. W. Thomas, and M. M. Humphries. 2005. Torpor patterns of hibernating eastern chipmunks Tamias striatus vary in response to the size and fatty acid composition of food hoards. Journal of Animal Ecology 74:692–700. Murphy, H. T., and J. Lovett-Doust. 2007. Accounting for regional niche variation in habitat suitability models. Oikos 116:99–110. Navas, C. A. 1996b. Implications of microhabitat selection and patterns of activity on the thermal ecology of high elevation neotropical anurans. Oecologia 108:617–626. —. 1996a. Metabolic physiology, locomotor performance, and thermal niche breadth in neotropical anurans. Physiological Zoology 69:1481–1501. —. 2006. Patterns of distribution of anurans in high Andean tropical elevations: insights from integrating biogeography and evolutionary physiology. Integrative and Comparative Biology 46:82–91. Navas, C. A., and C. Araujo. 2000. The use of agar models to study amphibian thermal ecology. Journal of Herpetology 34:330–334. Navas, C. A., R. S. James, J. M. Wakeling, K. M. Kemp, and I. A. Johnston. 1999. An integrative study of the temperature dependence of whole animal and muscle performance during jumping and swimming in the frog Rana temporaria. Journal of Comparative Physiology B 169:588–596. Neat, F., K. Fowler, V. French, and L. Partridge. 1995. Thermal evolution of growth efﬁciency in Drosophila melanogaster. Proceedings of the Royal Society B 260: 73–78. Nelson, K. C., and M. A. Palmer. 2007. Stream temperature surges under urbanization and climate change: data, models, and responses. Journal of the American Water Resources Association 43:440–452. New, M., M. Hulme, and P. Jones. 1999. Representing twentieth-century space-time climate variability. Part I: Development of a 1961–90 mean monthly terrestrial climatology. Journal of Climate 12:829–856. —. 2000. Representing twentieth-century space-time climate variability. Part II: Development of 1901–96 monthly grids of terrestrial surface climate. Journal of Climate 13:2217–2238. New, M., D. Lister, M. Hulme, and I. Makin. 2002. A highresolution data set of surface climate over global land areas. Climate Research 21:1–25. Newell, J. C., and T. P. Quinn. 2005. Behavioral thermoregulation by maturing adult sockeye salmon (Oncorhynchus nerka) in a stratiﬁed lake prior to spawning. Canadian Journal of Zoology 83:1232–1239.

REFERENCES

Newell, R. C., L. G. Johnson, and L. H. Kofoed. 1977. Adjustment of the components of energy balance in response to temperature change in Ostrea edulis. Oecologia 30:97–110. Newman, R. A. 1998. Ecological constraints on amphibian metamorphosis: interactions of temperature and larval density with responses to changing food level. Oecologia 115:9–16. Nice, C. C., and J. A. Fordyce. 2006. How caterpillars avoid overheating: behavioral and phenotypic plasticity of pipevine swallowtail larvae. Oecologia 146:541–548. Nicieza, A. G., L. Reiriz, and F. Brana. 1994. Variation in digestive performance between geographically disjunct populations of Atlantic salmon: countergradient in passage time and digestion rate. Oecologia 99: 243–251. Norry, F. M., O. A. Bubliy, and V. Loeschcke. 2001. Developmental time, body size and wing loading in Drosophila buzzatii from lowland and highland populations in Argentina. Hereditas 135:35–40. Norry, F. M., and V. Loeschcke. 2003. Heat-induced expression of a molecular chaperone decreases by selecting for long-lived individuals. Experimental Gerontology 38:673–681. Nosil, P., B. J. Crespi, C. P. Sandoval, and M. Kirkpatrick. 2006. Migration and the genetic covariance between habitat preference and performance. American Naturalist 167:E66-E78. Nowak, M. A., and K. Sigmund. 2004. Evolutionary dynamics of biological games. Science 303:793–799. Nussear, K. E., R. E. Espinoza, C. M. Gubbins, K. J. Field, and J. P. Hayes. 1998. Diet quality does not affect resting metabolic rate or body temperatures selected by an herbivorous lizard. Journal of Comparative Physiology B 168:183–189. O’Connor, M. P., and J. R. Spotila. 1992. Consider a spherical lizard: animals, models, and approximations. American Zoologist 32:179–193. O’Steen, S., and A. F. Bennett. 2003. Thermal acclimation effects differ between voluntary, maximum, and critical swimming velocities in two cyprinid ﬁshes. Physiological and Biochemical Zoology 76:484–496. Ohman, M. D., B. W. Frost, and E. B. Cohen. 1983. Reverse diel vertical migration: an escape from invertebrate predators. Science 220:1404–1407. Ohtsu, T., M. T. Kimura, and C. Katagiri. 1998. How Drosophila species acquire cold tolerance: qualitative changes of phospholipids. European Journal of Biochemistry 252:608–611. Ojanguren, A. F., and F. Braña. 2000. Thermal dependence of swimming endurance in juvenile brown trout. Journal of Fish Biology 56:1342–1347.

263

Oke, T. R. 1973. City size and the urban heat island. Atmospheric Environment 7:769–779. —. 1997. Urban climates and global change, Pages 273–287 in A. Perry, and R. Thompson, eds. Applied Climatology: Principles and Practice. London, Routledge. Oreskes, N., K. Shraderfrechette, and K. Belitz. 1994. Veriﬁcation, validation, and conﬁrmation of numerical models in the earth sciences. Science 263:641–646. Otterlei, E., G. Nyhammer, A. Folkvord, and S. O. Stefansson. 1999. Temperature- and size-dependent growth of larval and early juvenile Atlantic cod (Gadus morhua): a comparative study of Norwegian coastal cod and northeast Arctic cod. Canadian Journal of Fisheries and Aquatic Sciences 56:2099–2111. Ouedraogo, R. M., M. Cusson, M. S. Goettel, and J. Brodeur. 2003. Inhibition of fungal growth in thermoregulating locusts, Locusta migratoria, infected by the fungus Metarhizium anisopliae var acridum. Journal of Invertebrate Pathology 82:103–109. Ouedraogo, R. M., M. S. Goettel, and J. Brodeur. 2004. Behavioral thermoregulation in the migratory locust: a therapy to overcome fungal infection. Oecologia 138:312–319. Ouﬁero, C. E., and M. J. Angilletta. 2006. Convergent evolution of embryonic growth and development in the eastern fence lizard (Sceloporus undulatus). Evolution 60:1066–1075. Ouﬁero, C. E., A. J. Smith, and M. J. Angilletta. 2007. The importance of energetic versus pelvic constraints on reproductive allocation in the eastern fence lizard (Sceloporus undulatus). Biological Journal of the Linnean Society 91:513–521. Overgaard, J., J. G. Sorensen, S. O. Petersen, V. Loeschcke, and M. Holmstrup. 2005. Changes in membrane lipid composition following rapid cold hardening in Drosophila melanogaster. Journal of Insect Physiology 51:1173–1182. Packard, G. C., M. J. Packard, and L. L. McDaniel. 2001. Seasonal change in the capacity for supercooling by neonatal painted turtles. Journal of Experimental Biology 204:1667–1672. Padilla, D. K., and S. C. Adolph. 1996. Plastic inducible morphologies are not always adaptive: the importance of time delays in a stochastic environment. Evolutionary Ecology 10:105–117. Palaima, A. 2002. Tolerance to temperature: an experimental study of Daphnia (Crustacea: Cladocera). Ph.D. Dissertation, University of Miami, Coral Gables. —. 2007. The ﬁtness cost of generalization: present limitations and future possible solutions. Biological Journal of the Linnean Society 90:583–590.

264

REFERENCES

Palaima, A., and K. Spitze. 2004. Is a jack-of-alltemperatures a master of none? An experimental test with Daphnia pulicaria (Crustacea: Cladocera). Evolutionary Ecology Research 6:215–225. Parker, D. E. 2004. Large-scale warming is not urban. Nature 432:290–290. Parker, G. A., and M. Begon. 1986. Optimal egg size and clutch size: effects of environment and maternal phenotype. American Naturalist 128:573–592. Parker, S. L., and R. M. Andrews. 2007. Incubation temperature and phenotypic traits of Sceloporus undulatus: implications for the northern limits of distribution. Oecologia 151:218–231. Parmesan, C. 2006. Ecological and evolutionary responses to recent climate change. Annual Review of Ecology Evolution and Systematics 37:637–669. —. 2007. Inﬂuences of species, latitudes and methodologies on estimates of phenological response to global warming. Global Change Biology 13:1860–1872. Parmesan, C., S. Gaines, L. Gonzalez, D. M. Kaufman, J. Kingsolver, A. Townsend Peterson, and R. Sagarin. 2005. Empirical perspectives on species borders: from traditional biogeography to global change. Oikos 108:58–75. Parmesan, C., and H. Galbraith. 2004. Observed impacts of global climate change in the U.S., Pages 1–55. Arlington, Pew Center on Global Climate Change. Parmesan, C., N. Ryrholm, C. Stefanescu, J. K. Hill, C. D. Thomas, H. Descimon, B. Huntley et al. 1999. Poleward shifts in geographical ranges of butterﬂy species associated with regional warming. Nature 399:579–583. Parmesan, C., and G. Yohe. 2003. A globally coherent ﬁngerprint of climate change impacts across natural systems. Nature 421:37–42. Parry, D., R. A. Goyer, and G. J. Lenhard. 2001. Macrogeographic clines in fecundity, reproductive allocation, and offspring size of the forest tent caterpillar Malacosoma disstria. Ecological Entomology 26:281–291. Partridge, L., B. Barrie, K. Fowler, and V. French. 1994. Evolution and development of body size and cell size in Drosophila melanogaster in response to temperature. Evolution 48:1269–1276. Partridge, L., and J. A. Coyne. 1997. Bergmann’s rule in ectotherms: is it adaptive? Evolution 51:632–635. Parzefall, J. 2001. A review of morphological and behavioural changes in the cave molly, Poecilia mexicana, from Tabasco, Mexico. Environmental Biology of Fishes 62:263–275. Patino, S., J. Grace, and H. Banziger. 2000. Endothermy by ﬂowers of Rhizanthes lowii (Rafﬂesiaceae). Oecologia 124:149–155.

Pearcy, R. W. 1977. Acclimation of photosynthetic and respiratory carbon dioxide exchange to growth temperature in Atriplex lentiformis (Torr.) Wats. Plant Physiology 59:795–799. Pearson, R. G., and T. P. Dawson. 2003. Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecology and Biogeography 12:361–371. Pérez, E., F. Díaz, and S. Espina. 2003. Thermoregulatory behavior and critical thermal limits of the angelﬁsh Pterophyllum scalare (Lichtenstein) (Pisces: Cichlidae). Journal of Thermal Biology 28:531–537. Perrin, N. 1988. Why are offspring born larger when it is colder? Phenotypic plasticity for offspring size in the cladoceran Simocephalus vetulus (Müller). Functional Ecology 2:283–288. —. 1995. About Berrigan and Charnov’s life history puzzle. Oikos 73:137–139. Perrin, N., and J. F. Rubin. 1990. On dome-shaped reaction norms for age-to-size at maturity in ﬁshes. Functional Ecology 4:53–57. Perrin, N., and R. M. Sibly. 1993. Dynamic models of energy allocation and investment. Annual Review of Ecology and Systematics 24:379–410. Petavy, G., J. P. Morin, B. Moreteau, and J. R. David. 1997. Growth temperature and phenotypic plasticity in two Drosophila sibling species: probable adaptive changes in ﬂight capacities. Journal of Evolutionary Biology 10: 875–887. Petersen, A. M., T. T. Gleeson, and D. A. Scholnick. 2003. The effect of oxygen and adenosine on lizard thermoregulation. Physiological and Biochemical Zoology 76:339–347. Peterson, A. T., M. A. Ortega-Huerta, J. Bartley, V. SanchezCordero, J. Soberon, R. H. Buddemeier, and D. R. B. Stockwell. 2002. Future projections for Mexican faunas under global climate change scenarios. Nature 416: 626–629. Pfrender, M. E., and M. Lynch. 2000. Quantitative genetic variation in Daphnia: temporal changes in genetic architecture. Evolution 54:1502–1509. Phillips, P. K., and J. E. Heath. 2001. Heat loss in Dumbo: a theoretical approach. Journal of Thermal Biology 26: 117–120. Pianka, E. R. 1970. Comparative autecology of lizard Cnemidophorus tigris in different parts of its geographic range. Ecology 51:703–720. Piccione, G., G. Caola, and R. Reﬁnetti. 2002. Circadian modulation of starvation-induced hypothermia in sheep and goats. Chronobiology International 19: 531–541.

REFERENCES

Pigliucci, M. 1996. How organisms respond to environmental changes: from phenotypes to molecules (and vice versa). Trends in Ecology & Evolution 11:168–173. —. 2003. Phenotypic integration: studying the ecology and evolution of complex phenotypes. Ecology Letters 6:265–272. Pigliucci, M., and C. D. Schlichting. 1997. On the limits of quantitative genetics for the study of phenotypic evolution. Acta Biotheoretica 45:143–160. Pilon, J., and L. Santamaría. 2002. Clonal variation in the thermal response of the submerged aquatic macrophyte Potamogeton pectinatus. Journal of Ecology 90:141–152. Piper, S. C., and E. F. Stewart. 1996. A gridded global data set of daily temperature and precipitation for terrestrial biospheric modeling. Global Biogeochemical Cycles 10:757–782. Pires, H. H. R., C. R. Lazzari, P. E. Schilman, L. Diotaiuti, and M. G. Lorenzo. 2002. Dynamics of thermopreference in the Chagas disease vector Panstrongylus megistus (Hemiptera: Reduviidae). Journal of Medical Entomology 39:716–719. Platt, J. R. 1964. Strong inference. Science:347–353. Podrabsky, J. E., and G. N. Somero. 2006. Inducible heat tolerance in Antarctic notothenioid ﬁshes. Polar Biology 30:39–43. Polo, V., P. López, and J. Martín. 2005. Balancing the thermal costs and beneﬁts of refuge use to cope with persistent attacks from predators: a model and an experiment with an alpine lizard. Evolutionary Ecology Research 7:23–35. Porter, W. P., and D. M. Gates. 1969. Thermodynamic equilibria of animals with environment. Ecological Monographs 39:227–244. Porter, W. P., J. W. Mitchell, W. A. Beckman, and C. B. Dewitt. 1973. Behavioral implications of mechanistic ecology: thermal and behavioral modeling of desert ectotherms and their microenvironment. Oecologia 13:1–54. Pörtner, H. O. 2001. Climate change and temperaturedependent biogeography: oxygen limitation of thermal tolerance in animals. Naturwissenschaften 88:137–146. —. 2002. Climate variations and the physiological basis of temperature dependent biogeography: systemic to molecular hierarchy of thermal tolerance in animals. Comparative Biochemistry and Physiology A 132: 739–761. Pörtner, H. O., A. F. Bennett, F. Bozinovic, A. Clarke, M. A. Lardies, M. Lucassen, B. Pelster et al. 2006. Trade-offs in thermal adaptation: the need for a molecular to ecological integration. Physiological and Biochemical Zoology 79:295–313.

265

Pörtner, H. O., I. Hardewig, F. J. Sartoris, and P. L. M. Van Dijk. 1998. Energetic aspects of cold adaptation critical temperatures in metabolic, ionic and acid-base regulation?, Pages 88–120 in H. O. Pörtner, and R. Playle, eds. Cold Ocean Physiology Cambridge, Cambridge University Press. Pörtner, H. O., P. L. M. Van Dijk, I. Hardewig, and A. Sommer. 2000. Levels of metabolic cold adaptation: tradeoffs in eurythermal and stenothermal ectotherms, Pages 109–122 in W. Davison, C. Howard-Williams, and P. Broady, eds. Antarctic Ecosystems: Models for Wider Ecological Understanding. Christchurch, Caxton Press. Poulin, R. 2006. Global warming and temperaturemediated increases in cercarial emergence in trematode parasites. Parasitology 132:143–151. Powers, D. A., and P. M. Schulte. 1998. Evolutionary adaptations of gene structure and expression in natural populations in relation to a changing environment: a multidisciplinary approach to address the million-year saga of a small ﬁsh. Journal of Experimental Zoology 282:71–94. Prange, H. D., and G. B. Hamilton. 1992. Humidity selection by thermoregulating grasshoppers. Journal of Thermal Biology 17:353–355. Prasad, N. G., and M. N. Shakarad. 2004. Genetic correlations: transient truths of adaptive evolution. Journal of Genetics 83:3–6. Pravosudov, V. V., and J. R. Lucas. 2000. The costs of being cool: a dynamic model of nocturnal hypothermia by small food-caching birds in winter. Journal of Avian Biology 31:463–472. Present, T. M. C., and D. O. Conover. 1992. Physiological basis of latitudinal growth differences in Menidia menidia: variation in consumption or efﬁciency. Functional Ecology 6:23–31. Prinzinger, R., A. Pressmar, and E. Schleucher. 1991. Body temperature in birds. Comparative Biochemistry and Physiology A 99:499–506. Proulx, S. R., and T. Day. 2001. What can invasion analyses tell us about evolution under stochasticity in ﬁnite populations? Selection 2:1–15. Pulgar, J., F. Bozinovic, and F. P. Ojeda. 1999. Behavioral thermoregulation in the intertidal ﬁsh Girella laevifrons (Kyphosidae): the effect of starvation. Marine and Freshwater Behaviour and Physiology 32:27–38. Pulgar, J. M., M. Aldana, F. Bozinovic, and F. P. Ojeda. 2003. Does food quality inﬂuence thermoregulatory behavior in the intertidal ﬁsh Girella laevifrons? Journal of Thermal Biology 28:539–544. Pulliam, H. R. 2000. On the relationship between niche and distribution. Ecology Letters 3:349–361.

266

REFERENCES

Qualls, F. J., and R. Shine. 1998. Geographic variation in lizard phenotypes: importance of the incubation environment. Biological Journal of the Linnean Society 64:477–491. Quinn, J. F., and A. E. Dunham. 1983. On hypothesis testing in ecology and evolution. American Naturalist 122: 602–617. Rausch, R. N., L. I. Crawshaw, and H. L. Wallace. 2000. Effects of hypoxia, anoxia, and endogenous ethanol on thermoregulation in goldﬁsh, Carassius auratus. American Journal of Physiology 278:R545-R555. Ray, C. 1960. The application of Bergmann’s and Allen’s rules to the poikilotherms. Journal of Morphology 106:85–108. Reeve, M. W., K. Fowler, and L. Partridge. 2000. Increased body size confers greater ﬁtness at lower experimental temperature in male Drosophila melanogaster. Journal of Evolutionary Biology 13:836–844. Regal, P. J. 1966. Thermophilic response following feeding in certain reptiles. Copeia:588–590. —. 1971. Long term studies with operant conditioning techniques of temperature regulation patterns in reptiles. Journal de Physiologie 63:403–406. Rehfeldt, G. E., W. R. Wykoff, and C. C. Ying. 2001. Physiologic plasticity, evolution, and impacts of a changing climate on Pinus contorta. Climatic Change 50: 355–376. Reisen, W. K. 1995. Effect of temperature on Culex trasalis (Diptera: Culicidae) from the Coachella and San Joaquin Valleys of California. Journal of Medical Entomology 32:636–645. Renaud, J. M., and E. D. Stevens. 1983. The extent of longterm temperature compensation for jumping distance in the frog, Rana pipiens, and the toad, Bufo americanus. Canadian Journal of Zoology 61:1284–1287. Reyes, M. A., L. J. Corcuera, and L. Cardemil. 2003. Accumulation of HSP70 in Deschampsia antarctica Desv. leaves under thermal stress. Antarctic Science 15:345–352. Reynolds, W. W., and M. E. Casterlin. 1979. Behavioral thermoregulation and the ﬁnal preferendum paradigm. American Zoologist 19:211–224. Reznick, D., L. Nunney, and A. Tessier. 2000. Big houses, big cars, superﬂeas and the costs of reproduction. Trends in Ecology & Evolution 15:421–425. Rhen, T., and J. W. Lang. 1999. Temperature during embryonic and juvenile development inﬂuences growth in hatchling snapping turtles, Chelydra serpentina. Journal of Thermal Biology 24:33–41. Ricci, C. 1991. Comparison of ﬁve strains of a parthenogenetic species, Macrotrachela quadricornifera (Rotifera, Bdelloidea). Hydrobiologia 211:147–155.

Rice, S. 2004, Evolutionary Theory: Mathematical and Conceptual Foundations. Sunderland, Sinauer Associates, Inc. Riechert, S. E., and P. Hammerstein. 1983. Game theory in the ecological context. Annual Review of Ecology and Systematics 14:377–409. Riechert, S. E., and C. R. Tracy. 1975. Thermal balance and prey availability: bases for a model relating web-site characteristics to spider reproductive success. Ecology 56:265–284. Riley, P. A. 1997. Melanin. International Journal of Biochemistry & Cell Biology 29:1235–1239. Roberts, S. P. 2005. Effects of ﬂight behaviour on body temperature and kinematics during inter-male mate competition in the solitary desert bee Centris pallida. Physiological Entomology 30:151–157. Roberts, S. P., and M. E. Feder. 2000. Changing ﬁtness consequences of hsp70 copy number in transgenic Drosophila larvae undergoing natural thermal stress. Functional Ecology 14:353–357. Roberts, S. P., and J. F. Harrison. 1998. Mechanisms of thermoregulation in ﬂying bees. American Zoologist 38:492–502. Robinet, C., P. Baier, J. Pennerstorfer, A. Schopf, and A. Roques. 2007. Modelling the effects of climate change on the potential feeding activity of Thaumetopoea pityocampa (Den. & Schiff.) (Lep., Notodontidae) in France. Global Ecology and Biogeography 16:460–471. Robinson, S. J. W., and L. Partridge. 2001. Temperature and clinal variation in larval growth efﬁciency in Drosophila melanogaster. Journal of Evolutionary Biology 14:14–21. Robson, M. A., and D. B. Miles. 2000. Locomotor performance and dominance in male tree lizards, Urosaurus ornatus. Functional Ecology 14:338–344. Rock, J., R. M. Andrews, and A. Cree. 2000. Effects of reproductive condition, season, and site on selected temperatures of a viviparous gecko. Physiological and Biochemical Zoology 73:344–355. Roetzer, T., M. Wittenzeller, H. Haeckel, and J. Nekovar. 2000. Phenology in central Europe – differences and trends of spring phenophases in urban and rural areas. International Journal of Biometeorology 44:60–66. Roff, D. A. 1980. Optimizing development time in a seasonal environment: the ‘ups and downs’ of clinal variation. Oecologia 45:202–208. —. 1994. Optimality modeling and quantitative genetics: a comparison of the two approaches, Pages 49–66 in C. R. B. Boake, ed. Quantitative Genetic Studies of Behavioral Evolution. Chicago, University of Chicago Press. —. 2002, Life History Evolution. Sunderland, Sinauer Associates, Inc.

REFERENCES

Roff, D. A., and T. A. Mousseau. 1999. Does natural selection alter genetic architecture? An evaluation of quantitative genetic variation among populations of Allonemobius socius and A. fasciatus. Journal of Evolutionary Biology 12:361–369. Rogers, K. D., F. Seebacher, and M. B. Thompson. 2004. Biochemical acclimation of metabolic enzymes in response to lowered temperature in tadpoles of Limnodynastes peronii. Comparative Biochemistry and Physiology A 137:731–738. Rohmer, C., J. R. David, B. Moreteau, and D. Joly. 2004. Heat induced male sterility in Drosophila melanogaster: adaptive genetic variations among geographic populations and role of the Y chromosome. Journal of Experimental Biology 207:2735–2743. Romare, P., and L. A. Hansson. 2003. A behavioral cascade: top-predator induced behavioral shifts in planktivorous ﬁsh and zooplankton. Limnology and Oceanography 48:1956–1964. Rome, L. C., P. T. Loughna, and G. Goldspink. 1985. Temperature acclimation: improved sustained swimming performance in carp at low temperatures. Science 228: 194–196. Rome, L. C., E. D. Stevens, and H. B. John-Alder. 1992. The inﬂuence of temperature and thermal acclimation on physiological function, Pages 183–205 in M. E. Feder, and W. W. Burggren, eds. Environmental physiology of the Amphibians. Chicago, The University of Chicago Press. Ronce, O., and M. Kirkpatrick. 2001. When sources become sinks: migrational meltdown in heterogeneous habitats. Evolution 55:1520–1531. Root, T. L., J. T. Price, K. R. Hall, S. H. Schneider, C. Rosenzweig, and J. A. Pounds. 2003. Fingerprints of global warming on wild animals and plants. Nature 421:57–60. Rose, M. R., T. J. Nusbaum, and A. K. Chippindale. 1996. Laboratory evolution: the experimental Wonderland and the Cheshire Cat in M. R. Rose, and G. V. Lauder, eds. Adaptation. New York, Academic Press. Rosenheim, J. A. 2004. Top predators constrain the habitat selection games played by intermediate predators and their prey. Israel Journal of Zoology 50:129–138. Rossetti, Y., and M. Cabanac. 2006. Light versus temperature: an intersensitivity conﬂict in a gastropod (Lymnaea auricularia). Journal of Thermal Biology 31:514–520. Roth, M. 2002, Effects of cities on local climates Workshop for the IGES/APN Mega-City Project:1–13. Row, J. R., and G. Blouin-Demers. 2006. Thermal quality inﬂuences effectiveness of thermoregulation, habitat use, and behaviour in milk snakes. Oecologia 148: 1–11.

267

Rungruangsak-Torrissen, K., G. M. Pringle, R. Moss, and D. F. Houlihan. 1998. Effects of varying rearing temperatures on expression of different trypsin isozymes, feed conversion efﬁciency and growth in Atlantic salmon (Salmo salar L.). Fish Physiology and Biochemistry 19:247–255. Rustad, L. E., J. L. Campbell, G. M. Marion, R. J. Norby, M. J. Mitchell, A. E. Hartley, J. H. C. Cornelissen et al. 2001. A meta-analysis of the response of soil respiration, net nitrogen mineralization, and aboveground plant growth to experimental ecosystem warming. Oecologia 126: 543–562. Sack, R., A. Gochberg-Sarver, U. Rozovsky, M. Kedmi, S. Rosner, and A. Orr-Urtreger. 2005. Lower core body temperature and attenuated nicotine-induced hypothermic response in mice lacking the beta 4 neuronal nicotinic acetylcholine receptor subunit. Brain Research Bulletin 66:30–36. Sadowska, E. T., M. K. Labocha, K. Baliga, A. Stanisz, A. K. Wroblewska, W. Jagusiak, and P. Koteja. 2005. Genetic correlations between basal and maximum metabolic rates in a wild rodent: consequences for evolution of endothermy. Evolution 59:672–681. Saelim, S., and J. J. Zwiazek. 2000. Preservation of thermal stability of cell membranes and gas exchange in high temperature-acclimated Xylia xylocarpa seedlings. Journal of Plant Physiology 156:380–385. Samietz, J., S. Kroder, D. Schneider, and S. Dorn. 2006. Ambient temperature affects mechanosensory host location in a parasitic wasp. Journal of Comparative Physiology A 192:151–157. Samietz, J., M. A. Salser, and H. Dingle. 2005. Altitudinal variation in behavioural thermoregulation: local adaptation vs. plasticity in California grasshoppers. Journal of Evolutionary Biology 18:1087–1096. Sanborn, A. F., M. H. Villet, and P. K. Phillips. 2003. Hot-blooded singers: endothermy facilitates crepuscular signaling in African platypleurine cicadas (Hemiptera: Cicadidae: Platypleura spp.). Naturwissenschaften 90:305–308. —. 2004. Endothermy in African platypleurine cicadas: the inﬂuence of body size and habitat (Hemiptera: Cicadidae). Physiological and Biochemical Zoology 77: 816–823. Santamaría, L., and M. J. M. Hootsmans. 1998. The effect of temperature on the photosynthesis, growth and reproduction of a Mediterranean submerged macrophyte, Ruppia drepanensis. Aquatic Botany 60:169–188. Santamaría, L., and W. van Vierssen. 1997. Photosynthetic temperature responses of fresh- and brackishwater macrophytes: a review. Aquatic Botany 58: 135–150.

268

REFERENCES

Santos, M. 2007. Evolution of total net ﬁtness in thermal lines: Drosophila subobscura likes it ‘warm’. Journal of Evolutionary Biology 20:2361–2370. Santos, M., D. Brites, and H. Laayouni. 2006. Thermal evolution of pre-adult life history traits, geometric size and shape, and developmental stability in Drosophila subobscura. Journal of Evolutionary Biology 19:2006–2021. Sargent, R. C., P. D. Taylor, and M. R. Gross. 1987. Parental care and the evolution of egg size in ﬁshes. American Naturalist 129:32–46. Sarkar, S. 2004. From the Reaktionsnorm to the evolution of adaptive plasticity: a historical sketch, 1909–1999, Pages 10–30 in T. J. DeWitt, and S. M. Scheiner, eds. Phenotypic Plasticity: Functional and Conceptual Approaches. Oxford, Oxford University Press. Sarkar, S., and T. Fuller. 2003. Generalized norms of reaction for ecological developmental biology. Evolution & Development 5:106–115. Sartorius, S. S., J. P. S. do Amaral, R. D. Durtsche, C. M. Deen, and W. I. Lutterschmidt. 2002. Thermoregulatory accuracy, precision, and effectiveness in two sanddwelling lizards under mild environmental conditions. Canadian Journal of Zoology 80:1966–1976. Sarup, P., J. G. Sorensen, K. Dimitrov, J. S. F. Barker, and V. Loeschcke. 2006. Climatic adaptation of Drosophila buzzatii populations in southeast Australia. Heredity 96:479–486. Savage, V. M., J. F. Gillooly, J. H. Brown, G. B. West, and E. L. Charnov. 2004. Effects of body size and temperature on population growth. American Naturalist 163: E429–E441. Schaefer, J., and A. Ryan. 2006. Developmental plasticity in the thermal tolerance of zebraﬁsh Danio rerio. Journal of Fish Biology 69:722–734. Schäuble, C. S., and G. C. Grigg. 1998. Thermal ecology of the Australian agamid Pogona barbata. Oecologia 114:461–470. Scheiner, S. M. 1993. Genetics and evolution of phenotypic plasticity. Annual Review of Ecology and Systematics 24:35–68. —. 1998. The genetics of phenotypic plasticity. VII. Evolution in a spatially-structured environment. Journal of Evolutionary Biology 11:303–320. —. 2002. Selection experiments and the study of phenotypic plasticity. Journal of Evolutionary Biology 15: 889–898. Scheiner, S. M., and R. F. Lyman. 1991. The genetics of phenotypic plasticity. I. Heritability. Journal of Evolutionary Biology 2:95–107. Schlichting, C. D., and M. Pigliucci. 1998, Phenotypic Plasticity: A Reaction Norm Perspective. Sunderland, Sinauer Associates, Inc.

Schluter, D. 2000, The Ecology of Adaptive Radiation. Oxford, Oxford University Press. Schoenauer, M., and Z. Michalewicz. 1997. Evolutionary computation. Control and Cybernetics 26:307–338. Scholander, P. F., R. Hock, V. Walters, F. Johnson, and L. Irving. 1950a. Heat regulation in some arctic and tropical mammals and birds. Biological Bulletin 99: 237–258. Scholander, P. F., V. Walters, R. Hock, and L. Irving. 1950b. Body insulation of some Arctic and tropical mammals and birds. Biological Bulletin 99:225–236. Schoolﬁeld, R. M., P. J. H. Sharpe, and C. E. Magnuson. 1981. Non-linear regression of biological temperaturedependent rate models based on absolute reaction-rate theory. Journal of Theoretical Biology 88:719–731. Schulte, P. M. 2004. Changes in gene expression as biochemical adaptations to environmental change: a tribute to Peter Hochachka. Comparative Biochemistry and Physiology B 139:519–529. Schultz, E. T., K. E. Reynolds, and D. O. Conover. 1996. Countergradient variation in growth among newly hatched Fundulus heteroclitus: Geographic differences revealed by common-environment experiments. Functional Ecology 10:366–374. Schurmann, H., J. F. Steffensen, and J. P. Lomholt. 1991. The inﬂuence of hypoxia on the preferred temperature of rainbow trout Oncorhynchus mykiss. Journal of Experimental Biology 157:75–86. Schutze, M. K., and A. R. Clarke. 2008. Converse Bergmann cline in a Eucalyptus herbivore, Paropsis atomaria Olivier (Coleoptera: Chrysomelidae): phenotypic plasticity or local adaptation? Global Ecology and Biogeography 17:424–431. Scott, J. R., C. R. Tracy, and D. Pettus. 1982. A biophysical analysis of daily and seasonal utilization of climate space by a montane snake. Ecology 63:482–493. Scott, M., D. Berrigan, and A. A. Hoffmann. 1997. Costs and beneﬁts of acclimation to elevated temperature in Trichogramma carverae. Entomologia Experimentalis et Applicata 85:211–219. Scribner, S. J., and P. J. Weatherhead. 1995. Locomotion and antipredator behavior in three species of semiaquatic snakes. Canadian Journal of Zoology 73:321–329. Sears, M. W. 2006. Proceed with caution: invalidating tests of the cost–beneﬁt model of thermoregulation with spatially-explicit movement simulations. Integrative and Comparative Biology 46:E127. Sears, M. W., and M. J. Angilletta. 2007. Evaluating the costs of thermoregulation: simulating animal movements through spatially-structured environments deﬁne cost curves for small lizards. Integrative and Comparative Biology 47:E228.

REFERENCES

Sears, M. W., G. S. Bakken, M. J. Angilletta, and L. A. Fitzgerald. 2004. Using artiﬁcial neural networks to model the operative temperatures of small animals in a spatially-explicit context. Integrative and Comparative Biology 44:745. Seebacher, F., W. Davison, C. J. Lowe, and C. E. Franklin. 2005. A falsiﬁcation of the thermal specialization paradigm: compensation for elevated temperatures in Antarctic ﬁshes. Biology Letters 1:151–154. Seebacher, F., and G. C. Grigg. 1997. Patterns of body temperature in wild freshwater crocodiles, Crocodylus johnstoni: thermoregulation versus thermoconformity, seasonal acclimatization, and the effect of social interactions. Copeia:549–557. —. 2000. Social interactions compromise thermoregulation in crocodiles Crocodylus johnstoni and Crocodylus porosus Pages 310–316 in G. C. Grigg, F. Seebacher, and C. E. Franklin, eds. Crocodilian Biology and Evolution. Chipping Norton, Surrey Beatty and Sons Pty Ltd. —. 2001. Changes in heart rate are important for thermoregulation in the varanid lizard Varanus varius. Journal of Comparative Physiology B 171:395–400. Seebacher, F., G. C. Grigg, and L. A. Beard. 1999. Crocodiles as dinosaurs: behavioural thermoregulation in very large ectotherms leads to high and stable body temperatures. Journal of Experimental Biology 202:77–86. Seebacher, F., and R. Shine. 2004. Evaluating thermoregulation in reptiles: the fallacy of the inappropriately applied method. Physiological and Biochemical Zoology 77:688–695. Seebacher, F., and R. S. Wilson. 2006. Fighting ﬁt: thermal plasticity of metabolic function and ﬁghting success in the crayﬁsh Cherax destructor. Functional Ecology 20:1045–1053. Selong, J. H., T. E. McMahon, A. V. Zale, and F. T. Barrows. 2001. Effect of temperature on growth and survival of bull trout, with application of an improved method for determining thermal tolerance in ﬁshes. Transactions of the American Fisheries Society 130:1026–1037. Seymour, R. S., and A. J. Blaylock. 1999. Switching off the heater: inﬂuence of ambient temperature on thermoregulation by eastern skunk cabbageSymplocarpus foetidus. Journal of Experimental Botany 50:1525–1532. Sharpe, P. J. H., and D. W. DeMichele. 1977. Reaction kinetics of poikilotherm development. Journal of Theoretical Biology 64:649–670. Sherman, E., L. Baldwin, G. Fernandez, and E. Deurell. 1991. Fever and thermal tolerance in the toad Bufo marinus. Journal of Thermal Biology 16:297–301. Shine, R. 1980. Costs of reproduction in reptiles. Oecologia 46:92–100.

269

Shine, R., P. S. Harlow, M. J. Elphick, M. M. Olsson, and R. T. Mason. 2000. Conﬂicts between courtship and thermoregulation: the thermal ecology of amorous male garter snakes (Thamnophis sirtalis parietalis, Colubridae). Physiological and Biochemical Zoology 73:508–516. Shoemaker, V. H., L. L. McClanahan, P. C. Withers, S. S. Hillman, and R. C. Drewes. 1987. Thermoregulatory response to heat in the waterproof frogs Phyllomedusa and Chiromantis. Physiological Zoology 60:365–372. Shudo, E., P. Haccou, and Y. Iwasa. 2003. Optimal choice between feedforward and feedback control in gene expression to cope with unpredictable danger. Journal of Theoretical Biology 223:149–160. Sibly, R., and P. Calow. 1983. An integrated approach to life-cycle evolution using selective landscapes. Journal of Theoretical Biology 102:527–547. Sibly, R. M., and D. Atkinson. 1994. How rearing temperature affects optimal adult size in ectotherms. Functional Ecology 8:486–493. Sibly, R. M., D. Barker, J. Hone, and M. Pagel. 2007. On the stability of populations of mammals, birds, ﬁsh and insects. Ecology Letters 10:970–976. Sievert, L. M., and P. Andreadis. 1999. Speciﬁc dynamic action and postprandial thermophily in juvenile northern water snakes, Nerodia sipedon. Journal of Thermal Biology 24:51–55. Sievert, L. M., and V. H. Hutchison. 1988. Light versus heat: thermoregulatory behavior in a nocturnal lizard (Gekko gecko). Herpetologica 44:266–273. Sievert, L. M., D. M. Jones, and M. W. Puckett. 2005. Postprandial thermophily, transit rate, and digestive efﬁciency of juvenile cornsnakes, Pantherophis guttatus. Journal of Thermal Biology 30:354–359. Sigmund, K. 2005. John Maynard Smith and evolutionary game theory. Theoretical Population Biology 68:7–10. Sih, A. 1998. Game theory and predator-prey response races, Pages 221–238 in L. A. Dugatkin, and H. K. Reeve, eds. Game Theory and Animal Behavior. Oxford, Oxford University Press. Silbermann, R., and M. Tatar. 2000. Reproductive costs of heat shock protein in transgenic Drosophila melanogaster. Evolution 54:2038–2045. Sims, D. W., V. J. Wearmouth, E. J. Southall, J. M. Hill, P. Moore, K. Rawlinson, N. Hutchinson et al. 2006. Hunt warm, rest cool: bioenergetic strategy underlying diel vertical migration of a benthic shark. Journal of Animal Ecology 75:176–190. Sinclair, B. J. 1997. Seasonal variation in freezing tolerance of the New Zealand alpine cockroach Celatoblatta quinquemaculata. Ecological Entomology 22: 462–467.

270

REFERENCES

Sinclair, B. J., and S. P. Roberts. 2005. Acclimation, shock and hardening in the cold. Journal of Thermal Biology 30:557–562. Sinclair, B. J., P. Vernon, C. J. Klok, and S. L. Chown. 2003. Insects at low temperatures: an ecological perspective. Trends in Ecology & Evolution 18:257–262. Sinervo, B. 1990. The evolution of maternal investment in lizards: an experimental and comparative analysis of egg size and its effects on offspring performance. Evolution 44:279–294. Sinervo, B., P. Doughty, R. B. Huey, and K. Zamudio. 1992. Allometric engineering: a causal analysis of natural selection on offspring size. Science 258:1927–1930. Skelly, D. K. 2004. Microgeographic countergradient variation in the wood frog, Rana sylvetica. Evolution 58: 160–165. Skelly, D. K., L. N. Joseph, H. P. Possingham, L. K. Freidenburg, T. J. Farrugia, M. T. Kinnison, and A. P. Hendry. 2007. Evolutionary responses to climate change. Conservation Biology 21:1353–1355. Slatyer, R. O. 1977. Altitudinal variation in photosynthetic characteristics of snow gum, Eucalyptus pauciﬂorac Sieb. ex Spreng. IV. Temperature response of four populations grown at different temperatures. Australian Journal of Plant Physiology 4:583–594. Slip, D. J., and R. Shine. 1988. Thermophilic response to feeding of the diamond python, Morelia s. spilota (Serpentes, Boidae). Comparative Biochemistry and Physiology A 89:645–650. Smith, C. C., and S. D. Fretwell. 1974. Optimal balance between size and number of offspring. American Naturalist 108:499–506. Smith, E. M., and E. B. Hadley. 1974. Photosynthetic and respiratory acclimation to temperature inLedum groenlandicum populations. Arctic and Alpine Research 6:13–27. Smith, E. N. 1979. Behavioral and physiological thermoregulation of crocodilians. American Zoologist 19:239–247. Smith, G. R., and R. E. Ballinger. 1994. Thermal ecology of Sceloporus virgatus from southeastern Arizona, with comparison to Urosaurus ornatus. Journal of Herpetology 28:65–69. —. 1995. Temperature relationships of the tree lizard, Urosaurus ornatus, from desert and low-elevation montane populations in the southwestern USA. Journal of Herpetology 29:126–129. Smith, G. R., R. E. Ballinger, and J. D. Congdon. 1993. Thermal ecology of the high-altitude bunch grass lizard, Sceloporus scalaris. Canadian Journal of Zoology 71: 2152–2155. Smith, L. M., A. G. Appel, T. P. Mack, and G. J. Keever. 1999. Preferred temperature and relative humidity of males of

two sympatric Periplaneta cockroaches (Blattodea: Blattidae) denied access to water. Environmental Entomology 28:935–942. Snucins, E. J., and J. M. Gunn. 1995. Coping with a warm environment: behavioral thermoregulation by lake trout. Transactions of the American Fisheries Society 124:118–123. Snyder, G. K., and W. W. Weathers. 1975. Temperature adaptations in amphibians. American Naturalist 109:93–101. Soberón, J., and A. T. Peterson. 2005. Interpretation of models of fundamental ecological niches and species’ distributional areas. Biodiversity Informatics 2:1–10. Somero, G. N. 1995. Proteins and temperature. Annual Review of Physiology 57:43–68. —. 2000. University in diversity: a perspective on the methods, contributions, and future of comparative physiology. Annual Review of Physiology 62:927–937. —. 2002. Thermal physiology and vertical zonation of intertidal animals: optima, limits, and costs of living. Integrative and Comparative Biology 42:780–789. —. 2003. Protein adaptations to temperature and pressure: complementary roles of adaptive changes in amino acid sequence and internal milieu. Comparative Biochemistry and Physiology B 136:577–591. —. 2004. Adaptation of enzymes to temperature: searching for basic “strategies”. Comparative Biochemistry and Physiology B 139:321–333. Somero, G. N., E. Dahlhoff, and J. J. Lin. 1996. Stenotherms and eurytherms: mechanisms establishing thermal optima and tolerance ranges, Pages 53–78 in I. A. Johnston, and A. F. Bennett, eds. Animals and Temperature. Cambridge, Cambridge University Press. Somero, G. N., and P. W. Hochachka. 1971. Biochemical adaptation to environment. American Zoologist 11: 159–167. Sorci, G., J. Clobert, and S. Belichon. 1996. Phenotypic plasticity of growth and survival in the common lizard Lacerta vivipara. Journal of Animal Ecology 65:781–790. Sorensen, J. G., J. Dahlgaard, and V. Loeschcke. 2001. Genetic variation in thermal tolerance among natural populations of Drosophila buzzatii: down regulation of Hsp70 expression and variation in heat stress resistance traits. Functional Ecology 15:289–296. Sorensen, J. G., T. N. Kristensen, and V. Loeschcke. 2003. The evolutionary and ecological role of heat shock proteins. Ecology Letters 6:1025–1037. Sorensen, J. G., F. M. Norry, A. C. Scannapieco, and V. Loeschcke. 2005. Altitudinal variation for stress resistance traits and thermal adaptation in adult Drosophila buzzatii from the New World. Journal of Evolutionary Biology 18:829–837.

REFERENCES

Sorenson, J. G., and V. Loeschcke. 2002. Natural adaptation to environmental stress via physiological clockregulation of stress resistance in Drosophila. Ecology Letters 5:16–19. Soriano, P. J., A. Ruiz, and A. Arends. 2002. Physiological responses to ambient temperature manipulation by three species of bats from Andean cloud forests. Journal of Mammalogy 83:445–457. Spellerberg, I. F. 1972. Temperature tolerances of southeast Australian reptiles examined in relation to reptile thermoregulatory behaviour and distribution. Oecologia 9:23–46. Spotila, J. R., O. C. M. P., and G. S. Bakken. 1992. Biophysics of heat and mass transfer, Pages 59–80 in M. E. Feder, and W. W. Burggren, eds. Environmental Physiology of the Amphibians. Chicago, University of Chicago Press. Spotila, J. R., O. H. Soule, and D. M. Gates. 1972. The biophysical ecology of the alligator: heat energy budgets and climate spaces. Ecology 53:1094–1102. Stahlberg, F., M. Olsson, and T. Uller. 2001. Population divergence of developmental thermal optima in Swedish common frogs, Rana temporaria. Journal of Evolutionary Biology 14:755–762. Stamper, J. L., I. Zucker, D. A. Lewis, and J. Dark. 1998. Torpor in lactating Siberian hamsters subjected to glucoprivation. American Journal of Physiology 43:R46-R51. Stanwell-Smith, D., and L. S. Peck. 1998. Temperature and embryonic development in relation to spawning and ﬁeld occurrence of larvae of three Antarctic echinoderms. Biological Bulletin 194:44–52. Stapley, J. 2006. Individual variation in preferred body temperature covaries with social behaviours and colour in male lizards. Journal of Thermal Biology 31:362–369. Stearns, S. C. 1989. The evolutionary signiﬁcance of phenotypic plasticity. Bioscience 39:436–445. —. 1992, The Evolution of Life Histories. Oxford, Oxford University Press. —. 2000. Life history evolution: successes, limitations, and prospects Naturwissenschaften 87:476 486. Steigenga, M. J., B. J. Zwaan, P. M. Brakeﬁeld, and K. Fischer. 2005. The evolutionary genetics of egg size plasticity in a butterﬂy. Journal of Evolutionary Biology 18:281–289. Stelzer, C. P. 2002. Phenotypic plasticity of body size at different temperatures in a planktonic rotifer: mechanisms and adaptive signiﬁcance. Functional Ecology 16:835–841. Stenseng, E., C. E. Braby, and G. N. Somero. 2005. Evolutionary and acclimation-induced variation in the thermal limits of heart function in congeneric marine snails (genus Tegula): implications for vertical zonation. Biological Bulletin 208:138–144.

271

Stephens, D. W., J. S. Brown, and R. C. Ydenberg. 2007. Foraging: Behavior and Ecology, Pages 576. Chicago, University Of Chicago Press. Stephens, D. W., and J. R. Krebs. 1986, Foraging Theory. Princeton, Princeton University Press. Steppan, S. J., P. C. Phillips, and D. Houle. 2002. Comparative quantitative genetics: evolution of the G matrix. Trends in Ecology & Evolution 17:320–327. Stevenson, R. D. 1985. Body size and limits to the daily range of body temperature in terrestrial ectotherms. American Naturalist 125:102–117. Stevenson, R. D., C. R. Peterson, and J. S. Tsuji. 1985. The thermal dependence of locomotion, tongue ﬂicking, digestion, and oxygen consumption in the wandering garter snake. Physiological Zoology 58:46–57. Stillman, J. H. 2003. Acclimation capacity underlies susceptibility to climate change. Science 301:65. Stillman, J. H., and G. N. Somero. 2000. A comparative analysis of the upper thermal tolerance limits of eastern Paciﬁc porcelain crabs, genus Petrolisthes: Inﬂuences of latitude, vertical zonation, acclimation, and phylogeny. Physiological and Biochemical Zoology 73:200–208. Stillwell, R. C., and C. W. Fox. 2005. Complex patterns of phenotypic plasticity: interactive effects of temperature during rearing and oviposition. Ecology 86:924–934. Stoks, R., M. De Block, F. Van de Meutter, and F. Johansson. 2005. Predation cost of rapid growth: behavioural coupling and physiological decoupling. Journal of Animal Ecology 74:708–715. Strong, K. W., and G. R. Daborn. 1980. The inﬂuence of temperature on energy budget variables, body size, and seasonal occurrence of the isopod Idotea baltica (Pallas). Canadian Journal of Zoology 58:1992–1996. Sun, H. J., and E. I. Friedmann. 2005. Communities adjust their temperature optima by shifting producer-toconsumer ratio, shown in lichens as models: II. Experimental veriﬁcation. Microbial Ecology 49:528–535. Sutherst, R. W., G. F. Maywald, and A. S. Bourne. 2007. Including species interactions in risk assessments for global change. Global Change Biology 13:1843–1859. Sykes, M. T., I. C. Prentice, and W. Cramer. 1996. A bioclimatic model for the potential distributions of north European tree species under present and future climates. Journal of Biogeography 23:203–233. Talloen, W., H. Van Dyck, and L. Lens. 2004. The cost of melanization: butterﬂy wing coloration under environmental stress. Evolution 58:360–366. Tansey, M. R., and T. D. Brock. 1972. The upper temperature limit for eukaryotic organisms. Proceedings of the National Academy of Sciences 69:2426–2428. Tatar, M. 1999. Transgenes in the analysis of life span and ﬁtness. American Naturalist 154:S67-S81.

272

REFERENCES

—. 2000. Transgenic organisms in evolutionary ecology. Trends in Ecology & Evolution 15:207–211. Taylor, P. D., and G. C. Williams. 1984. Demographic parameters at evolutionary equilibrium. Canadian Journal of Zoology 62:2264–2271. Temple, G. K., and I. A. Johnston. 1998. Testing hypotheses concerning the phenotypic plasticity of escape performance in ﬁsh of the family Cottidae. Journal of Experimental Biology 201:317–331. Templeton, A. R. 1986. Coadaptation and outbreeding depression, Pages 105–121 in M. E. Soulé, ed. Conservation Biology: The Science of Scarcity and Diversity. Sunderland, Sinauer Associates, Inc. Ten Hwang, Y., S. Lariviere, and F. Messier. 2007. Energetic consequences and ecological signiﬁcance of heterothermy and social thermoregulation in striped skunks (Mephitis mephitis). Physiological and Biochemical Zoology 80:138–145. Terblanche, J. S., and S. L. Chown. 2006. The relative contributions of developmental plasticity and adult acclimation to physiological variation in the tsetse ﬂy, Glossina pallidipes (Diptera, Glossinidae). Journal of Experimental Biology 209:1064–1073. Thomas, C. D., A. Cameron, R. E. Green, M. Bakkenes, L. J. Beaumont, Y. C. Collingham, B. F. N. Erasmus et al. 2004. Extinction risk from climate change. Nature 427: 145–148. Thomson, L. J., M. Robinson, and A. A. Hoffmann. 2001. Field and laboratory evidence for acclimation without costs in an egg parasitoid. Functional Ecology 15: 217–221. Torti, V. M., and P. O. Dunn. 2005. Variable effects of climate change on six species of North American birds. Oecologia 145:486–495. Tracy, C. R. 1976. A model of the dynamic exchanges of water and energy between a terrestrial amphibian and its environment. Ecological Monographs 46:293–326. Tracy, C. R., and K. A. Christian. 1986. Ecological relations among space, time, and thermal niche axes. Ecology 67:609–615. —. 2005. Preferred temperature correlates with evaporative water loss in hylid frogs from northern Australia. Physiological and Biochemical Zoology 78:839–846. Tracy, C. R., F. H. Vanberkum, J. S. Tsuji, R. D. Stevenson, J. A. Nelson, B. M. Barnes, and R. B. Huey. 1984. Errors resulting from linear approximations in energy balance equations. Journal of Thermal Biology 9:261–264. Travis, J., and D. N. Reznick. 1998. Experimental approaches to the study of evolution, Pages 437–459 in W. J. Resetarits Jr., and J. Bernardo, eds. Experimental Ecology: Issues and Perspectives. Oxford, Oxford University Press.

Tsai, M. L., C. F. Dai, and H. H. Chen. 1998. Desiccation resistance of two semiterrestrial isopods, Ligia exotica and Ligia taiwanensis (Crustacea) in Taiwan. Comparative Biochemistry and Physiology A 119: 361–367. Tsai, T. S., and M. C. Tu. 2005. Postprandial thermophily of Chinese green tree vipers, Trimeresurus s. stejnegeri: interfering factors on snake temperature selection in a thigmothermal gradient. Journal of Thermal Biology 30:423–430. Tsuji, J. S. 1988. Thermal acclimation of metabolism in Sceloporus lizards from different latitudes. Physiological Zoology 61:241–253. Tu, M. C., and V. H. Hutchison. 1994. Inﬂuence of pregnancy on thermoregulation of water snakes (Nerodia rhombifera). Journal of Thermal Biology 19:255–259. Tveiten, H., S. E. Solevåg, and H. K. Johnsen. 2001. Holding temperature during the breeding season inﬂuences ﬁnal maturation and egg quality in common wolfﬁsh. Journal of Fish Biology 58:374–385. Underwood, B. A. 1991. Thermoregulation and energetic decision-making by the honeybees Apis cerana, Apis dorsata and Apis laboriosa. Journal of Experimental Biology 157:19–34. van’t Land, J., P. van Putten, B. Zwaan, A. Kamping, and W. van Delden. 1999. Latitudinal variation in wild populations of Drosophila melanogaster: heritabilities and reaction norms. Journal of Evolutionary Biology 12:222–232. van Berkum, F. H. 1986. Evolutionary patterns of the thermal sensitivity of sprint speed in Anolis lizards. Evolution 40:594–604. —. 1988. Latitudinal patterns of the thermal sensitivity of sprint speed in lizards. American Naturalist: 327–343. van Berkum, F. H., R. B. Huey, and B. A. Adams. 1986. Physiological consequences of thermoregulation in a tropical lizard (Ameiva festiva). Physiological Zoology 59:464–472. Van Damme, R., D. Bauwens, A. M. Castilla, and R. F. Verheyen. 1989. Altitudinal variation of the thermal biology and running performance in the lizard Podarcis tiliguerta. Oecologia 80:516–524. —. 1990. Comparative thermal ecology of the sympatric lizards Podarcis tiliguerta and Podarcis sicula. Acta Oecologica 11:503–512. Van Damme, R., D. Bauwens, and R. F. Verheyen. 1991. The thermal dependence of feeding behavior, food consumption and gut passage time in the lizard Lacerta vivipara Jacquin. Functional Ecology 5:507–517. van der Have, T. M. 2002. A proximate model for thermal tolerance in ectotherms. Oikos 98:141–155.

REFERENCES

van Dijk, P. L. M., G. Staaks, and I. Hardewig. 2002. The effect of fasting and refeeding on temperature preference, activity and growth of roach, Rutilus rutilus. Oecologia 130:496–504. Van Doorslaer, W., and R. Stoks. 2005a. Growth rate plasticity to temperature in two damselﬂy species differing in latitude: contributions of behaviour and physiology. Oikos:599–605. van Doorslaer, W., and R. Stoks. 2005b. Thermal reaction norms in two Coenagrion damselﬂy species: contrasting embryonic and larval life-history traits. Freshwater Biology 50:1982–1990. Van Doorslaer, W., R. Stoks, E. Jeppesen, and L. De Meester. 2007. Adaptive microevolutionary responses to simulated global warming in Simocephalus vetulus: a mesocosm study. Global Change Biology 13: 878–886. van Huis, A., P. W. Arendse, M. Schilthuizen, P. P. Wiegers, H. Heering, M. Hulshof, and N. K. Kaashoek. 1994. Uscana lariophaga, egg parasitoid of bruchid beetle storage pests of cowpea in West Africa: the effect of temperature and humidity. Entomologia experimentalis et applicata 70:41–53. van Tienderen, P. H. 1991. Evolution of generalists and specialists in spatially heterogeneous environments. Evolution 45:1317–1331. Vannote, R. L., and B. W. Sweeney. 1980. Geographic analysis of thermal equilibria: a conceptual model for evaluating the effect of natural and modiﬁed thermal regimes on aquatic insect communities. American Naturalist 115:667–695. Vargas, R. I., W. A. Walsh, E. B. Jang, J. W. Armstrong, and D. T. Kanehisa. 1996. Survival and development of immature stages of four Hawaiian fruit ﬂies (Diptera: Tephritidae) reared at ﬁve constant temperatures. Annals of the Entomological Society of America 89:64–69. Vázquez, D. P., and R. D. Stevens. 2004. The latitudinal gradient in niche breadth: concepts and evidence. American Naturalist 164:E1-E19. Verdu, J. R., A. Diaz, and E. Galante. 2004. Thermoregulatory strategies in two closely related sympatricScarabaeus species (Coleoptera: Scarabaeinae). Physiological Entomology 29:32–38. Vernberg, F. J. 1962. Comparative physiology: latitudinal effects on physiological properties of animal populations. Annual Review of Physiology 24:517–544 Via, S. 1987. Genetic constraints on the evolution of phenotypic plasticity, Pages 47–71 in V. Loeschcke, ed. Genetic Constraints on Adaptive Evolution. Berlin, SpringerVerlag.

273

—. 1994. The evolution of phenotypic plasticity: what do we really know?, Pages 35–57 in L. A. Real, ed. Ecological Genetics. Princeton, Princeton University Press. Via, S., R. Gomulkiewicz, G. de Jong, S. M. Scheiner, C. D. Schlichting, and P. H. van Tienderen. 1995. Adaptive phenotypic plasticity: consensus and controversy. Trends in Ecology & Evolution 10:212–217. Via, S., and R. Lande. 1985. Genotype-environment interaction and the evolution of phenotypic plasticity. Evolution 39:505–522. Visser, M. E., and C. Both. 2005. Shifts in phenology due to global climate change: the need for a yardstick. Proceedings of the Royal Society B 272:2561–2569. Visser, M. E., L. J. M. Holleman, and P. Gienapp. 2006. Shifts in caterpillar biomass phenology due to climate change and its impact on the breeding biology of an insectivorous bird. Oecologia 147:164–172. Vitt, L. J., and S. S. Sartorius. 1999. HOBOs, Tidbits and lizard models: the utility of electronic devices in ﬁeld studies of ectotherm thermoregulation. Functional Ecology 13:670–674. Vona, V., V. D. Rigano, O. Lobosco, S. Carfagna, S. Esposito, and C. Rigano. 2004. Temperature responses of growth, photosynthesis, respiration and NADH: nitrate reductase in crophilic and mesophilic algae. New Phytologist 163:325–331. Waldschmidt, S. R., S. M. Jones, and W. P. Porter. 1986. The effect of body temperature and feeding regime on activity, passage time, and digestive coefﬁcient in the lizard Uta stansburiana. Physiological Zoology 59: 376–383. Wallman, H. L., and W. A. Bennett. 2006. Effects of parturition and feeding on thermal preference of Atlantic stingray,Dasyatis sabina (Lesueur). Environmental Biology of Fishes 75:259–267. Walsberg, G. E., and B. O. Wolf. 1996. A test of the accuracy of operative temperature thermometers for studies of small ectotherms. Journal of Thermal Biology 21:275–281. Walther, G. R., S. Berger, and M. T. Sykes. 2005. An ecological ‘footprint’ of climate change. Proceedings of the Royal Society B 272:1427–1432. Walther, G. R., E. S. Gritti, S. Berger, T. Hickler, Z. Y. Tang, and M. T. Sykes. 2007. Palms tracking climate change. Global Ecology and Biogeography 16:801–809. Walther, G. R., E. Post, P. Convey, A. Menzel, C. Parmesan, T. J. C. Beebee, J. M. Fromentin et al. 2002. Ecological responses to recent climate change. Nature 416:389–395. Walton, B. M., and A. F. Bennett. 1993. Temperature dependent color change in Kenyan chameleons. Physiological Zoology 66:270–287.

274

REFERENCES

Walvoord, M. E. 2003. Cricket frogs maintain body hydration and temperature near levels allowing maximum jump performance. Physiological and Biochemical Zoology 76:825–835. Warner, R. R. 1980. The coevolution of behavioral and life history characteristics, Pages 151–188 in G. W. Barlow, and J. Silverberg, eds. Sociobiology: Beyond Nature/Nurture? Boulder, Westview Press. Warren, C. E., and G. E. Davis. 1967. Laboratory studies on the feeding, bioenergetics, and growth of ﬁsh, Pages 175–214 in S. D. Gerking, ed. The Biological Basis of Freshwater Fish Production. Oxford, Blackwell Scientiﬁc Publications. Warren, M., M. A. McGeoch, S. W. Nicolson, and S. L. Chown. 2006. Body size patterns in Drosophila inhabiting a mesocosm: interactive effects of spatial variation in temperature and abundance. Oecologia 149: 245–255. Warwick, C. 1991. Observations on disease-associated preferred body temperatures in reptiles. Applied Animal Behaviour Science 28:375–380. Webb, J. K., and M. J. Whiting. 2005. Why don’t small snakes bask? Juvenile broad-headed snakes trade thermal beneﬁts for safety. Oikos 110:515–522. Weetman, D., and D. Atkinson. 2002. Antipredator reaction norms for life history traits in Daphnia pulex: dependence on temperature and food. Oikos 98:299–307. Weng, Q. H., D. S. Lu, and J. Schubring. 2004. Estimation of land surface temperature-vegetation abundance relationship for urban heat island studies. Remote Sensing of Environment 89:467–483. Werner, Y. L. 1973. Optimal temperatures for inner-ear performance in gekkonoid lizards. Journal of Experimental Zoology 195:319–352. —. 1990. Do gravid females of oviparous gekkonid. lizards maintain elevated body temperatures? Hemidactylus frenatus and Lepidodactylus lugubris on Oahu. Amphibia-Reptilia 11:200–204. West-Eberhard, M. J. 2003, Developmental Plasticity and Evolution. Oxford, Oxford University Press. Westman, W., and F. Geiser. 2004. The effect of metabolic fuel availability on thermoregulation and torpor in a marsupial hibernator. Journal of Comparative Physiology B 174:49–57. Whitaker, B. D., and K. L. Poff. 1980. Thermal adaptation of thermosensing and negative thermotaxis in Dictyostelium. Experimental Cell Research 128:87–93. White, M. A., R. R. Nemani, P. E. Thornton, and S. W. Running. 2002. Satellite evidence of phenological differences between urbanized and rural areas of the eastern United States deciduous broadleaf forest. Ecosystems 5: 260–273.

Whitlock, M. C., P. C. Phillips, F. B. G. Moore, and S. J. Tonsor. 1995. Multiple ﬁtness peaks and epistasis. Annual Review of Ecology and Systematics 26:601–629. Widdows, J., and B. L. Bayne. 1971. Temperature acclimation of Mytilus edulis with reference to its energy budget. Journal of the Marine Biological Association of the United Kingdom 51:827–843. Wiens, J. J., and C. H. Graham. 2005. Niche conservatism: integrating evolution, ecology, and conservation biology. Annual Review of Ecology and Systematics 36:519–539. Wiens, J. J., C. H. Graham, D. S. Moen, S. A. Smith, and T. W. Reeder. 2006. Evolutionary and ecological causes of the latitudinal diversity gradient in hylid frogs: treefrog trees unearth the roots of high tropical diversity. American Naturalist 168:579–596. Wieser, W. 1994. Cost of growth in cells and organisms: general rules and comparative aspects. Biological Reviews 69:1–33. Wiggins, P. R., and P. B. Frappell. 2000. The inﬂuence of haemoglobin on behavioural thermoregulation and oxygen consumption in Daphnia carinata. Physiological and Biochemical Zoology 73:153–160. Wildhaber, M. L. 2001. The trade-off between food and temperature in the habitat choice of bluegill sunﬁsh. Journal of Fish Biology 58:1476–1478. Wildhaber, M. L., and P. J. Lamberson. 2004. Importance of the habitat choice behavior assumed when modeling the effects of food and temperature on ﬁsh populations. Ecological Modelling 175:395–409. Wilhelm, F. M., and D. W. Schindler. 2000. Reproductive strategies of Gammarus lacustris (Crustacea: Amphipoda) along an elevational gradient. Functional Ecology 14. Williams, J. W., S. T. Jackson, and J. E. Kutzbacht. 2007. Projected distributions of novel and disappearing climates by 2100 AD. Proceedings of the National Academy of Sciences 104:5738–5742. Wills, C. A., and S. J. Beaupre. 2000. An application of randomization for detecting evidence of thermoregulation in timber rattlesnakes (Crotalus horridus) from northwest Arkansas. Physiological and Biochemical Zoology 73:325–334. Wilson, R. S. 2001. Geographic variation in thermal sensitivity of jumping performance in the frog Limnodynastes peronii. Journal of Experimental Biology 204:4227–4236. —. 2005. Temperature inﬂuences the coercive mating and swimming performance of male eastern mosquitoﬁsh. Animal Behaviour 70:1387–1394. Wilson, R. S., and C. E. Franklin. 1999. Thermal acclimation of locomotor performance in tadpoles of the frog Limnodynastes peronii. Journal of Comparative Physiology B 169:445–451.

REFERENCES

—. 2000a. Effect of ontogenetic increases in body size on burst swimming performance in tadpoles of the striped marsh frog, Limnodynastes peronii. Physiological and Biochemical Zoology 73:142–152. —. 2000b. Inability of adult Limnodynastes peronii (Amphibia: Anura) to thermally acclimate locomotor performance. Comparative Biochemistry and Physiology A 127:21–28. —. 2002. Testing the beneﬁcial acclimation hypothesis. Trends in Ecology & Evolution 17:66–70. Wilson, R. S., C. E. Franklin, W. Davison, and P. Kraft. 2001. Stenotherms at sub-zero temperatures: thermal dependence of swimming performance in Antarctic ﬁsh. Journal of Comparative Physiology B 171:263–269. Wilson, R. S., E. Hammill, and I. A. Johnston. 2007. Competition moderates the beneﬁts of thermal acclimation to reproductive performance in male eastern mosquitoﬁsh. Proceedings of the Royal Society B 274:1199–1204. Wilson, R. S., R. S. James, and I. A. Johnston. 2000. Thermal acclimation of locomotor performance in tadpoles and adults of the aquatic frog Xenopus laevis. Journal of Comparative Physiology B 170:117–124. Windig, J. J., C. G. F. de Kovel, and G. de Jong. 2004. Genetics and mechanics of plasticity, Pages 31–49 in T. J. DeWitt, and S. M. Scheiner, eds. Phenotypic Plasticity: Functional and Conceptual Approaches. Oxford, Oxford University Press. Winne, C. T., and M. B. Keck. 2005. Intraspeciﬁc differences in thermal tolerance of the diamondback watersnake (Nerodia rhombifer): effects of ontogeny, latitude, and sex. Comparative Biochemistry and Physiology A 140:141–149. Withers, P. C., and J. D. Campbell. 1985. Effects of environmental cost on thermoregulation in the desert iguana. Physiological Zoology 58:329–339. Witters, L. R., and L. Sievert. 2001. Feeding causes thermophily in the woodhouse’s toad (Bufo woodhousii). Journal of Thermal Biology 26:205–208. Wolf, J. B., and E. D. Brodie. 1998. The coadaptation of parental and offspring characters. Evolution 52: 299–308. Woods, H. A. 1999. Egg-mass size and cell size: effects of temperature on oxygen distribution. American Zoologist 39:244–252. Woodward, F. I. 1990. The impact of low temperatures in controlling the geographical distribution of plants. Philosophical Transactions of the Royal Society B 326:585–593. Wu, B. J., P. L. Else, L. H. Storlien, and A. J. Hulbert. 2001. Molecular activity of Na+/K+-ATPase from different sources is related to the packing of membrane lipids. Journal of Experimental Biology 204:4271–4280.

275

Wu, B. J., A. J. Hulbert, L. H. Storlien, and P. L. Else. 2004. Membrane lipids and sodium pumps of cattle and crocodiles: an experimental test of the membrane pacemaker theory of metabolism. American Journal of Physiology 287:R633–R641. Xiong, F. S. S., E. C. Mueller, and T. A. Day. 2000. Photosynthetic and respiratory acclimation and growth response of Antarctic vascular plants to contrasting temperature regimes. American Journal of Botany 87:700–710. Yadav, J. P., and B. N. Singh. 2005. Evolutionary genetics of Drosophila ananassae III. Effect of temperature on certain ﬁtness trait’s in two natural populations. Journal of Thermal Biology 30:457–466. Yamahira, K., and D. O. Conover. 2002. Intra- vs. interspeciﬁc latitudinal variation in growth: adaptation to temperature or seasonality? Ecology 83:1252–1262. Yamahira, K., M. Kawajiri, K. Takeshi, and T. Irie. 2007. Inter- and intrapopulation variation in thermal reaction norms for growth rate: evolution of latitudinal compensation in ectotherms with a genetic constraint. Evolution 61:1577–1589. Yamasaki, T., T. Yamakawa, Y. Yamane, H. Koike, K. Satoh, and S. Katoh. 2002. Temperature acclimation of photosynthesis and related changes in photosystem II electron transport in winter wheat. Plant Physiology 128: 1087–1097. Yamori, W., K. Suzuki, K. Noguchi, M. Nakai, and I. Terashima. 2006. Effects of Rubisco kinetics and Rubisco activation state on the temperature dependence of the photosynthetic rate in spinach leaves from contrasting growth temperatures. Plant Cell and Environment 29:1659–1670. Yampolsky, L. Y., and S. M. Scheiner. 1996. Why larger offspring at lower temperatures? A demographic approach. American Naturalist 147:86–100. Yang, J., Y. Y. Sun, H. An, and X. Ji. 2008. Northern grass lizards (Takydromus septentrionalis) from different populations do not differ in thermal preference and thermal tolerance when acclimated under identical thermal conditions. Journal of Comparative Physiology B 178:343–349. Zamudio, K. R., R. B. Huey, and W. D. Crill. 1995. Bigger isn’t always better: body size, developmental and parental temperature and male territorial success in Drosophila melanogaster. Animal Behaviour 49:671–677. Zari, T. A. 1998. Effects of sexual condition on food consumption and temperature selection in the herbivorous desert lizard, Uromastyx philbyi. Journal of Arid Environments 38:371–377. Zbikowska, E. 2004. Does behavioural fever occur in snails parasitised with trematode larvae? Journal of Thermal Biology 29:675–679.

276

REFERENCES

Zeilstra, I., and K. Fischer. 2005. Cold tolerance in relation to developmental and adult temperature in a butterﬂy. Physiological Entomology 30:92–95. Zera, A. J., and L. G. Harshman. 2001. The physiology of life history trade-offs in animals. Annual Review of Ecology and Systematics 32:95–126. Zhang, Y. P., and X. A. Ji. 2004. The thermal dependence of food assimilation and locomotor performance in southern grass lizards, Takydromus sexlineatus (Lacertidae). Journal of Thermal Biology 29:45–53.

Zhao, C. Y., Z. R. Nan, and G. D. Cheng. 2005. Methods for modelling of temporal and spatial distribution of air temperature at landscape scale in the southern Qilian mountains, China. Ecological Modelling 189: 209–220. Zhou, X. H., X. Z. Liu, L. L. Wallace, and Y. Q. Luo. 2007. Photosynthetic and respiratory acclimation to experimental warming for four species in a tallgrass prairie ecosystem. Journal of Integrative Plant Biology 49:270–281.

Author Index

Ackermann, M. 210, 211 Addo-Bediako, A. 23, 27, 58 Adolph, S. C. 18, 29, 32, 96, 145 Alderdice, D. F. 146 Alford, R. A. 116 Almansa, E. 44 Alonzo, S. H. 120, 206 Ambruster, W. S. 97 Ames, D. 122 Amo, L. 114 Anderson, A. R. 75, 85 Anderson, D. R. 6, 38, 40 Anderson, J. L. 92 Andersson, S. 67 Andreadis, P. 103 Andrews, R. M. 101, 220 Angilletta, M. J. 1, 40, 56, 62, 66, 76–8, 90, 110, 112, 121, 127, 154, 160–1, 165, 166, 168–70, 177, 186, 192–5 Araujo, C. 31 Araujo, M. B. 219 Archer, M. A. 10 Armbruster, P. 172 Arnett, A. E. 66 Arnﬁeld, A. J. 216 Arnold, S. J. 6, 8, 9, 15, 79, 92 Artacho, P. 196 Asbury, D. A. 96, 192 Ashton, K. G. 159, 160 Atkin, O. K. 128 Atkinson, D. 158, 159, 162, 165, 166, 169, 170, 171, 179 Ayers, D. Y. 96 Ayres, M. P. 56 Ayrinhac, A. 61 Azevedo, R. B. R. 172, 173, 179 Baer, C. F. 85 Baird, O. E. 88 Bakken, G. S. 20, 21, 31, 32, 117, 198 Balanya, J. 225 Barber, B. J. 92 Barlow, C. A. 35 Bartelt, P. E. 31 Bartholomew, G. A. 96, 97

Bartlett, P. N. 29, 30 Barton, N. H. 233–5 Basheer, I. A. 33 Bashey, F. 29, 32 Baur, B. 172 Bauwens, D. 37, 96, 112, 188, 189 Bayne, B. L. 76, 139 Beard, L. 125 Beaugrand, G. 217 Beaupre, S. J. 29, 109 Becker, U. 71 Bedford, G. 30, 32 Begon, M. 176, 177 Beitinger, T. L. 88, 111, 187, 220 Belk, M. C. 67, 159 Bell, G. 226 Belliure, J. 118 Bemis, W. E. 2 Bennett, A. F. 2, 14, 55, 71–2, 75, 94, 97, 119, 121, 134, 146, 181, 186, 188, 196 Bennett, W. A. 92 Berrigan, D. 6, 38, 48, 127, 130, 135, 163–7, 169, 179 Berry, J. 67, 128, 146 Bertalanffy, von L. 163, 169 Berven, K. A. 66, 67, 71 Beuchat, C. A. 106–7, 190 Bicego, K. C. 88 Bilcke, J. 130 Billerbeck, J. M. 76, 78 Bishop, J. A. 97 Björkman, O. 67, 128, 146 Black-Samuelsson, S. 67 Blackburn, T. M. 159 Blanckenhorn, W. U. 159, 161, 171, 172 Blaylock, A. J. 123 Block, B. A. 97 Blouin-Demers, G. 103, 110 Bochdanovits, Z. 162, 179 Boily, P. 121 Boivin, G. 75, 85 Bossert, W. H. 195 Both, C. 217

Bowker, R. G. 92 Bowler, K. 1, 97, 126 Bradshaw, D. 116 Bradshaw, W. E. 23, 225 Brattstrom, B. H. 88 Brock, T. D. 35, 56 Brodie, E. D. 181, 196, 198 Bronikowski, A. M. 95 Brown, G. P. 32, 103 Brown, J. H. 1, 2 Bubli, O. A. 85 Buchner, O. 139 Buckley, L. B. 111, 220, 221, 224 Bull, C. M. 116 Bull, J. J. 73, 85 Burger, R. 79, 230, 231–2 Burggren, W. W. 2 Burnham, K. P. 6, 38, 40 Cabanac, M. 92 Calow, P. 174 Campbell, G. S. 22 Campbell, J. D. 108, 110 Carlson, S. M. 196 Carrascal, L. M. 118 Carrière, Y. 75, 85 Carter, P. A. 2, 56 Casey, T. M. 30, 98 Castañeda, L. E. 58 Castenholz, R. W. 96 Casterlin, M. E. 92 Cerda, X. 209 Chamberlin, T. C. 6 Chapin, F. S. 71 Chapin, M. C. 71 Chappell, M. A. 22, 96 Charnov, E. L. 166, 179 Chen, X. J. 192 Chintalapati, S. 136 Chown, S. L. 58, 97, 127, 134, 149 Christian, K. 5, 30, 32, 41, 89, 90, 111, 116 Ciais, P. 218 Cichoñ, M. 179 Clark, C. W. 120 277

278

AU T H O R I N D E X

Clarke, A. 27, 56, 119 Clarke, A. R. 161 Cleavitt, N. 71 Clusella-Trullas, S. 97 Coelho, J. R. 96 Cohen, M. P. 116 Collins, S. 226 Congdon, J. D. 177 Connell, J. H. 111 Conner, J. K. 14 Conover, D. O. 53, 67, 78, 179, 196 Cooper, B. S. 57 Cooper, V. S. 72, 95 Cooper, W. E. 97 Cortemeglia, C. 220 Cossins, A. R. 1, 97 Costanzo, J. P. 136 Coyne, J. A. 160, 161 Crawford, E. C. 92 Crill, W. D. 128 Crockett, E. L. 136 Crompton, A. W. 119 Crowder, L. B. 116 Crowley, S. R. 116 Crozier, L. 220, 221 Cunningham, S. C. 67, 139, 148, 149 Czesak, M. E. 171

Daborn, G. R. 167 Danks, H. V. 88 Dark, J. 123 David, J. R. 9, 69, 159 David, P. 3 Davis, A. J. 220 Davison, I. R. 67 Dawidowicz, P. 115, 199 Dawkins, R. 200 Dawson, T. P. 219 Dawson, W. R. 187 Day, T. 200, 211 Dayan, T. 160, 208 De Block, M. 158 De Boer, M. K. 67 de Jong, G. 9, 152, 162, 179 de Macario, E. C. 136 de Mazancourt, C. 236 de Quieroz, A. 160 DeAngelis, D. L. 200 Debat, V. 3 Deere, J. A. 149 DeMichele, D. W. 40, 46 Demont, M. 161 Denny, M. W. 29 Deutsch, C. A. 224, 226 DeVries, A. L. 136

DeWitt, T. J. 3, 7, 135, 143 Diaz, F. 187 Diaz, J. A. 29 Diego-Rasilla, F. J. 105, 115 Dill, L. M. 7, 203 Dillon, M. E. 161 Doebeli, M. 210, 211 Dohm, M. R. 92, 116 Dorcas, M. E. 32 Downes, S. 112, 115, 118 Drent, J. 56 Dreyer, B. S. 56 Du, W. G. 171, 192 Duellman, W. E. 22, 88 Duever, M. J. 146 Dufresne, F. 159 Dullinger, S. 220 Dunham, A. E. 6, 18, 29, 32, 89, 158, 168–70, 195 Dunn, P. O. 217 Dwyer, G. 220, 221 Dybdahl, M. F. 159 Dzialowski, E. M. 21, 31, 32, 96–7, 112 Ebert, D. 54, 70 Edmunds, P. J. 220 Ehleringer, J. 96, 116 Elith, J. 219 Ellison, A. M. 6 Ellison, G. T. H. 123 Ellner, S. 106–7, 190 Else, P. L. 44, 76, 121 Endler, J. A. 15 Eppley, R. W. 94 Falconer, D. S. 9, 11 Farmer, C. G. 119–21 Feder, M. E. 2, 101, 136–9, 151 Feldman, C. R. 159, 160 Felsenstein, J. 17, 18 Fiedler, K. 158, 170, 171 Fields, P. A. 41, 43, 45, 75 Finnerty, J. R. 97 Fischer, K. 127, 158, 168, 170, 171, 172, 173 Fisher, R. 8 Fitter, A. H. 216 Fitter, R. S. R. 216 Fitzpatrick, L. C. 88, 187 Fleming, I. A. 172 Folk, D. G. 85 Forchhammer, M. C. 216 Fordyce, J. A. 98 Forseth, I. 96, 116 Fox, C. W. 158, 159, 171

Franco, A. M. A. 217 Franklin, C. E. 126, 127, 130, 136 Frappell, P. B. 105 Frazier, M. 94, 194 Freidenburg, L. K. 92 Fretwell, S. D. 174, 201 Frid, L. 106 Friedmann, E. I. 67 Fry, F. E. J. 146 Fuller, T. 4

Gabriel, W. 47–51, 53–7, 61, 76, 81–5, 135, 141–5, 154, 155, 157, 182–6, 194, 196, 229 Gadgil, M. 195 Gagliano, M. 176 Galbraith, H. 216, 217, 224–5 Galen, C. 116 Garland, T. 2, 17, 18, 56, 119, 121, 188 Gaston, K. J. 27, 58 Gates, D. M. 19, 21–3, 27, 29–31, 113, 117 Gavrilets, S. 152, 207, 208, 209 Gebhardt, M. D. 159 Geiser, F. 122, 123 Gerald, G. W. 116 Ghalambor, C. K. 23, 27, 29, 58 Gibbons, J. W. 177 Gilbert, P. 9, 10, 61 Gilchrist, G. W. 8, 12, 35, 36, 48, 51–3, 73, 75, 76, 81, 83–5, 95, 128, 145, 162, 190, 194, 195, 196, 211, 232 Gill, D. E. 66 Gillespie, J. 8, 10 Gilligan, M. R. 159 Gillooly, J. F. 94, 192 Gilman, S. E. 29, 220 Gittleman, J. L. 18 Glanville, E. J. 146, 186 Gomes, F. R. 121 Gomi, T. 225 Gomulkiewicz, R. 10, 79–80 Good, D. S. 92, 197 Gordo, O. 218 Gotelli, N. J. 66 Gotthard, K. 3, 77 Graham, C. H. 219 Grand, T. C. 202, 203 Grant, B. W. 29, 30, 32, 89, 109 Greenberg, N. 97 Greenwald, O. E. 55 Grether, G. F. 54 Grewal, P. S. 69 Grifﬁths, J. S. 146 Grigg, G. C. 29, 30, 32, 96, 97, 111, 122, 123, 125

AU T H O R I N D E X

Grimm, N. B. 216 Gross, M. R. 172 Guderley, H. 44 Guégan, J. F. 33 Guisan, A. 219 Gunn, J. M. 96 Gurevitch, J. 160 Gvozdík, L. 116, 117

Hadley, E. B. 146, 148, 184 Hajmeer, M. 33 Hallas, R. 61 Hamilton, G. B. 97 Hammerstein, P. 200 Hansen, T. F. 10, 18 Hansson, L. A. 205 Harley, C. D. G. 29 Harrington, R. 217 Harrison, J. F. 88, 97 Harshman, L. G. 77 Hart, J. S. 146 Hartnoll, R. G. 97 Harvell, C. D. 220 Harvey, P. H. 18 Hawkins, A. J. S. 76 Hayes, J. P. 119, 121 Haynie, D. T. 19 Hazel, J. R. 43, 136 Hazel, W. N. 97, 126 Heath, J. E. 88–9, 96, 121 Hedges, L. V. 160 Heino, M. 6 Heinrich, B. 88, 96–8, 122, 123 Helmuth, B. S. T. 29, 30, 32 Hendry, A. P. 176 Henwood, K. 29 Hebert, P. D. N. 159 Hercus, M. J. 137 Herczeg, G. 101–2 Hereford, J. 15 Hermisson, J. 209 Hernandez, R. M. 154 Hero, J. M. 71, 159, 172 Herr, D. 97 Hertz, A. 13 Hertz, P. E. 29, 32, 47, 58, 62, 75, 90, 93, 109, 110, 187, 196, 227 Hijmans, R. J. 219 Hilborn, R. 2, 5 Hirshﬁeld, M. F. 195 Hochochka, P. W. 2, 17, 35, 41, 75, 94, 136 Hoefer, A. M. 118 Hoffmann, A. A. 57, 61, 66, 126, 134, 135, 138 Hofmann, G. E. 32, 136

Holder, K. K. 73, 85 Holland, K. N. 97 Holt, R. D. 8, 86, 225, 226 Holzapfel, C. M. 23, 225 Houston, D. D. 159 Howe, P. D. 32, 33 Huey, R. B. 1, 12, 18, 36–8, 47, 48, 61, 65, 70, 72, 75, 85, 88, 89, 95–103, 105, 109–12, 122, 127, 128, 130, 131, 135, 139, 151, 162, 169, 181, 186, 188–92, 196, 201, 202, 226, 231, 232 Hughes, L. 216 Hughes, N. F. 202, 203 Hugie, D. M. 203 Hulbert, A. J. 76 Humphries, M. M. 123, 228 Hung, H. W. 36 Husak, J. F. 55 Hutchinson, G. E. 219 Hutchinson, V. H. 38, 92, 127

Imsland, A. K. 67, 76 Intergovernmental Panel on Climate Change (IPCC) 214, 215, 218, 224 Izem, R. 74, 75, 85

Jaenicke, R. 35, 41, 76 James, A. C. 73, 95, 162, 179 Jayne, B. C. 55 Jeffery, W. R. 97 Jiang, L. 220 Jin, Y. M. 136 Joern, A. 209 Johansson, J. 232–3 John-Alder, H. B. 62 Johnsen, O. C. 127 Johnson, J. B. 6 Johnson, T. P. 146 Johnston, I. A. 126 Jonassen, T. M. 67, 76 Jones, A. G. 85, 237 Jonsson, B. 76 Jump, A. S. 225, 226

Kaitala, V. 7 Kalberer, S. R. 127 Kamil, A. C. 2 Kammer, A. E. 96 Kane, S. L. 159 Karl, I. 168 Karlsson, B. 69, 171 Kassen, R. 70

279

Kauffman, A. S. 123 Kauffman, J. S. 134 Kawecki, T. J. 54, 70 Kearney, M. 30, 220 Keck, M. B. 58 Kelty, J. D. 134, 139 Kemp, D. J. 97, 98 Kemp, T. S. 119 Kerr, G. D. 116 Kerr, J. T. 25 Kessler, K. 116 Ketterson, E. D. 14 Kiefer, M. C. 110 Kindlmann, P. A. 166, 168–9 Kingsolver, J. G. 8, 10, 15, 22, 29, 41, 48, 65, 74, 75, 77, 79, 81, 85, 97, 98, 139, 151, 231, 232 Kirkpatrick, M. 10, 79–80, 85, 233–5 Kirschbaum, M. U. F. 218 Klok, C. J. 44, 58 Kluger, M. J. 105, 106 Knies, J. L. 73–4 Kobler, D. 13 Koella, J. C. 163–4, 165, 167 Kopp, M. 207, 209 Koteja, P. 119, 121 Kozłowski, J. 7, 48, 163–5, 167, 171, 179 Krebs, J. R. 2, 176 Krebs, R. A. 85, 137, 138–9, 151 Kreuger, B. 96 Krochmal, A. R. 198 Krockenberger, A. K. 97, 98 Kronfeld-Schor, N. 208 Krueger, C. C. 88 Ksiazek, A. 121 Kulczycki, A. 220

Lacey, E. P. 97 Ladyman, M. 116 Lagerspetz, K. Y. H. 88, 126 Lakoff, G. 5 Lamb, R. J. 66 Lamberson, P. J. 116 Lampert, W. 115, 116, 202, 203 Lande, R. 8, 9, 15, 80, 152, 181, 229–30, 231 Lang, J. W. 128 Langerhans, R. B. 143 Lankford, T. E. 56 Laptikhovsky, V. 172 Larimore, R. W. 146 Laugen, A. T. 66, 161 Laurila, A. 76 Law, R. 195 Le Galliard, J. F. 92

280

AU T H O R I N D E X

Lee, R. E. 134, 136, 139 Lek, S. 33 Lenormand, T. 8 Lenski, R. E. 14, 71, 72, 75 Leroi, A. M. 127 Levins, R. 5, 35, 47 Levinton, J. S. 66, 67, 76 Licht, P. 58, 90, 92, 154 Lillywhite, H. B. 97 Lima, F. P. 217, 219 Lima, S. L. 7, 199 Lindgren, B. 76 Loeschcke, V. 85, 126, 127, 134, 137, 139, 149, 151, 154, 159 Lonsdale, D. J. 66, 67, 76 Loose, C. J. 115, 199 Lopez, P. 115 Lorenzon, P. 116 Lovegrove, B. G. 122, 123 Lovett-Doust, J. 219 Lucas, H. 201 Lucas, J. R. 124 Luh, H. K. 18 Luo, L. G. 192 Luoto, M. 219 Lustick, S. I. 96 Lutterschmidt, W. I. 38, 103 Lyman, R. F. 8, 9 Lynch, J. F. 101 Lynch, M. 8–11, 47, 49–51, 53–7, 61, 76, 79, 81–2, 84, 85, 135, 141–3, 155, 157, 182, 183, 194, 229–32

Mac, M. J. 103 Macario, A. J. L. 136 McCabe, J. 162, 179 McCullough, E. C. 23 McGinley, M. A. 174, 178, 212, 213 McKechnie, A. E. 122, 123, 124 McNab, B. K. 127 McPeek, M. A. 77 Magnum, C. P. 2, 17 Magnuson, J. J. 111, 117 Maness, J. D. 127 Mangel, M. 2, 5, 120 Martin, J. 115 Martin, T. L. 190–1, 192 Martins, E. P. 17, 18 Marx, J. C. 41 Masaki, S. 170 Maynard Smith, J. 6, 7, 199, 200 Measey, G. J. 159 Medvick, P. A. 111, 114, 199 Meffe, G. K. 172 Meiri, S. 160

Mejia-Ortiz, L. M. 97 Mendéndez, R. 217 Merila, J. 56, 66 Merrill, R. M. 220 Michalewicz, Z. 10 Michod, R. E. 195 Miles, D. B. 18, 55 Miller, K. 58 Miller, S. R. 96 Misra, R. K. 162 Mitchell, J. W. 19 Mitchell, M. 10 Mitchell, W. A. 6, 7 Mitten, J. B. 42 Moller, H. 158 Mondal, S. 36 Mongold, J. A. 95 Monteith, J. J. 27 Mooij, W. M. 200 Mooney, H. A. 146, 148 Moore, J. A. 56 Moran, N. A. 143 Morin, P. J. 220 Morrison, C. 71, 159, 172 Mosseau, T. A. 10 Munoz, J. L. P. 96 Munro, D. 123 Murphy, H. T. 219 Myers, J. H. 106

Nadeau, P. 110 Navas, C. A. 31, 62, 97 Neat, F. 179 Nelson, K. C. 216 Neuner, G. 139 New, M. 25, 26, 27 Newell, J. C. 89 Newell, R. C. 146 Newman, R. A. 159 Nice, C. C. 98 Nicieza, A. G. 76 Nicolson, S. W. 127, 134 Niehaus, A. C. 154 Nolan, V. 14 Norry, F. M. 66, 137 Nosil, P. 186, 196 Nowak, M. A. 7, 199 Núñez, R. E. 5 Nussear, K. E. 103 Nylin, S. 3

O’Connor, M. P. 30, 96–7, 112 Ohman, M. D. 116 Oke, T. R. 216, 225 Omland, K. S. 6

Oreskes, N. 6 Ostrovsky, M. 25 Otterlei, E. 56 Ouedraogo, R. M. 105–6, 199 Ouﬁero, C. E. 66, 76, 177 Overgaard, J. 136

Packard, G. C. 58, 126 Padilla, D. K. 145 Palaima, A. 75 Palmer, M. A. 216 Parker, D. E. 216 Parker, G. A. 176, 177 Parker, S. L. 220 Parmesan, C. 27, 216, 217, 224–6 Parry, D. 172 Partridge, L. 73, 76, 95, 160–2, 179 Parzefall, J. 97 Patino, S. 97 Pearson, R. G. 219 Penuelas, J. 225, 226 Pérez, E. 187 Perez-Mellado, V. 105 Perrin, N. 164, 165, 167, 169, 174, 178–9 Petavy, G. 159 Petersen, A. M. 105 Peterson, A. T. 218, 219, 220 Peterson, C. R. 31, 32 Pfrender, M. E. 10 Phillips, P. K. 121 Pianka, E. R. 110 Piccione, G. 123 Pigliucci, M. 3, 6, 41, 180 Pilon, J. 68 Piper, S. C. 23, 25 Platt, J. R. 6 Podrabsky, J. E. 139 Poff, K. L. 88 Polo, V. 114–15, 199 Porter, W. P. 19, 22, 23, 29, 30, 220 Pörtner, H. O. 35, 44, 45 Post, E. 216 Potter, D. A. 96 Poulin, R. 220 Powers, D. A. 42 Prange, H. D. 97 Prasad, N. G. 121 Pravosudov, V. V. 124 Present, T. M. C. 67, 78 Price, G. R. 199 Prinzinger, R. 88, 97 Proulx, S. R. 200 Pulgar, J. 92, 104 Pulliam, H. R. 220 Purvis, A. 18

AU T H O R I N D E X

Qualls, F. J. 66 Quinn, J. F. 6 Quinn, T. P. 89

Raboud, C. 172 Rai, U. 36 Rausch, R. N. 105 Ray, C. 158 Read, J. 67, 139, 148, 149 Reeve, E. C. R. 162 Reeve, M. W. 162, 179 Regal, P. J. 103, 112 Rehfeldt, G. E. 233 Reinert, H. K. 103 Reisen, W. K. 56 Reyes, M. A. 136 Reynolds, W. W. 92 Reznick, D. N. 14, 77 Rhen, T. 128 Rice, S. 8 Riechert, S. E. 116, 200 Riley, P. A. 97 Roberts, S. P. 97, 118, 126, 134, 137 Robinet, C. 220 Robinson, S. J. W. 76, 179 Robson, M. A. 55 Roetzer, T. 216 Roff, D. A. 10, 55, 158 Rogers, K. D. 136 Rohmer, C. 69 Romare, P. 205 Rome, L. C. 35 Ronce, O. 85 Root, T. L. 216 Rose, M. R. 16 Rosenheim, J. A. 205 Rossetti, Y. 92 Roth, M. 216 Rothery, P. 119 Roughgarden, J. 111 Row, J. R. 110 Ruben, J. A. 121 Rubin, J. F. 165 Rungruangsak-Torrissen, K. 42, 43 Rustad, L. E. 218 Ryan, A. 154

Sack, R. 198 Sadowska, E. T. 121 Samietz, J. 36, 96 Sanborn, A. F. 98, 105 Santamaría, L. 68 Santos, M. 95, 161, 162 Sanz, J. J. 218 Sargent, R. C. 176

Sarkar, S. 3, 4 Sartorius, S. S. 32 Savage, V. M. 1, 94, 192 Schaefer, J. 154 Schäuble, C. S. 32 Scheiner, S. M. 3, 6, 8, 9, 11, 14, 80, 152, 174, 179 Schindler, D. W. 172 Schlichting, C. D. 3, 6 Schluter, D. 70 Schoenauer, M. 10 Scholander, P. F. 97 Schoolﬁeld, R. M. 40, 46 Schulte, P. M. 42, 136 Schultz, E. T. 53, 67 Schurmann, H. 105 Schutze, M. K. 161 Scott, J. R. 22 Scott, M. 137 Scriber, J. M. 56 Seamons, T. R. 196 Sears, M. W. 32, 33, 111, 112, 121, 228 Seebacher, F. 29, 30, 96, 97, 111, 146, 186 Seymour, R. S. 123 Shakarad, M. N. 121 Sharpe, P. J. H. 40, 46 Sherman, E. 105 Shine, R. 66, 96, 117, 118 Shoemaker, V. H. 97 Shudo, E. 139 Sibly, R. 174 Sibly, R. M. 48, 162, 164, 165, 169, 171, 174, 179 Sievert, L. M. 92, 103 Sigmund, K. 7, 199 Sih, A. 201, 203, 204, 205 Silbermann, R. 137 Sims, D. W. 116 Sinclair, B. J. 126, 134 Sinervo, B. 14, 176 Singh, B. N. 159 Skelly, D. K. 66, 92, 225 Skinner, J. D. 123 Slatkin, M. 96, 98–102, 105, 113, 122, 201 Slatyer, R. O. 146, 148 Smith, C. C. 174 Smith, E. M. 146, 148, 184 Smith, E. N. 96 Smith, L. M. 116 Snucins, E. J. 96 Snyder, G. K. 58 Soberón, J. 219, 220 Somero, G. N. 1, 15, 35, 41, 42, 44, 58, 75, 76, 94, 136–9 Sorci, G. 71

281

Sorensen, J. G. 61, 126, 127, 134, 136, 139, 149, 154 Soriano, P. J. 123 Spellerberg, I. F. 58 Spezzano, L. C. 116 Spitze, K. 75 Spotila, J. R. 19, 22, 30 Stahlberg, F. 56, 67 Stamper, J. L. 123 Stapley, J. 112 Stearns, S. C. 3, 48, 158, 159, 163, 195 Steigenga, M. J. 173 Stelzer, C. P. 171, 179 Stephens, D. W. 2, 176 Steppan, S. J. 10, 13, 81 Stevens, R. D. 27 Stevenson, R. D. 1, 36, 37, 97, 192 Stewart, E. F. 23, 25 Stillman, J. H. 58, 152, 226 Stillwell, R. C. 158, 159 Stoks, R. 67, 76, 158 Storm, J. J. 154 Strong, K. W. 167 Sun, H. J. 67 Sutherst, R. W. 219 Sweeney, B. W. 27 Sykes, M. T. 220 Talloen, W. 97 Tansey, M. R. 35 Targett, T. E. 56 Tatar, M. 14, 137 Taylor, C. E. 10 Taylor, P. D. 174, 175 Templeton, A. R. 181 Ten Hwang, Y. 124 Terblanche, J. S. 127 Thomas, C. D. 217 Thomson, L. J. 137 Thuiller, W. 219 Tinkle, D. W. 195 Torti, V. M. 217 Tracy, C. R. 20, 21, 29, 30, 55, 111, 116, 117 Travis, J. 14, 85 Trueb, L. 22, 88 Tsai, M. L. 97 Tveiten, H. 171 Underwood, B. A. 104 Unsworth, M. H. 27 Vainio, L. A. 88 Valone, T. J. 6, 7 van Berkum, F. H. 37, 64, 92 Van Damme, R. 29, 30, 88, 89, 110, 192

282

AU T H O R I N D E X

van der Have, T. M. 36, 41, 46, 162 Van Dongen, S. 159 Van Doorslaer, W. 67, 76, 226 Van Dyck, H. 69, 171 van Tienderen, P. H. 151 van Vierssen, W. 68 Vannote, R. L. 27 van’t Land, J. 179 Vázquez, D. P. 27 Verdu, J. R. 88 Vernberg, F. J. 56 Via, S. 8, 9, 13, 80, 151, 152 Visser, M. E. 217, 218 Vitt, L. J. 32 Wallman, H. L. 92 Walsberg, G. E. 31 Walsh, B. 8, 9, 11 Walther, G. R. 215, 216, 219 Walton, B. M. 97 Walvoord, M. E. 97 Warner, R. R. 196 Warren, M. 161 Warwick, C. 106

Weatherhead, P. J. 32, 103 Weathers, W. W. 58 Weavers, B. W. 32, 90 Webb, J. K. 114 Weetman, D. 159 Weng, Q. H. 216 Werner, Y. L. 36, 90 West-Eberhard, M. J. 3, 9 Westman, W. 123 Whitaker, B. D. 88 Whiting, M. J. 112 Whitlock, M. C. 181 Whitman, D. W. 22, 96 Widdows, J. 139 Wiens, J. J. 1, 219 Wieser, W. 77, 179 Wiggins, P. R. 105 Wiklund, C. 69 Wildhaber, M. L. 116 Wilhelm, F. M. 172 Williams, E. E. 43 Williams, G. C. 174, 175 Williams, J. W. 237 Wilson, R. S. 56, 62, 126, 127, 130, 136, 154

Windig, J. J. 11 Winne, C. T. 58 Withers, P. C. 108, 110 Wolf, B. O. 31 Wolf, J. B. 181, 196 Woods, H. A. 41, 79, 176 Woodward, F. I. 225–6 Wu, B. J. 44

Yadav, J. P. 159 Yamahira, K. 67, 78, 85 Yamasaki, T. 136 Yamori, W. 136 Yampolsky, L. Y. 174, 179 Yang, J. 187 Yohe, G. 216, 217

Zamudio, K. R. 128 Zbikowska, E. 106 Zeilstra, I. 127 Zera, A. J. 77 Zhao, C. Y. 25

Species Index

Acer saccharum 147 Acyrthosiphon pisum 66 Aglais urticae 90 Ambrosia psilostachya 129 Amphibolurus 60 Anolis 62 carolinensis 104 cristatellus 93 gundlachi 60, 93 krugi 60 semilineatus 60 Antaresia childreni 107 Antheraea assama 98 Aporia crataegi 220 Arabidopsis thaliana 66 Arianta arbustorum 172 Aster ericoides 129 Atalopedes campestris 221, 222 Atelopus 64 Atriplex glabriuscula 131 lentiformis 131, 147 sabulosa 131

Bactrocera cucurbitae 39 dorsalis 39 latifrons 39 Ballia callitricha 129 Barbus barbus 132 Bicyclus anynana 172–3 Bufo americanus 133 marinus 104 woodhousii 104

Carassius auratus 132 Carex aquatilis 71 Carlina vulgaris 71 Cenchrus ciliaris 129 Centris pallida 118 Ceratitis capitata 39 Chalcides ocellatus 107 Charina bottae 104, 107

Cherax destructor 132 dispar 132 Chlorella sacchrophila 39 sorokiniana 39 Christianus marmoratus 90 Chrysosporium pannorum 225 Cladina 67 Coenagrion hastulatem 39 puella 39 Colestethus 64 Collisella scabra 219–20 Colobanthus quitensis 129 Crocodylus porosus 134 Crotalus viridis 107 Culex tarsalis 57 Cyprinus carpio 132

Danio rerio 154 Daphnia 116, 203 magna 39, 56, 67, 68, 115 pulex 159 pulicaria 203 Dasyatis sabina 104 Deschampsia antarctica 129 Desmognathus fuscus 104 Dipsosaurus dorsalis 22, 41, 108 Drosophila 58–61, 62, 150, 161 buzzatii 59, 61, 137, 149, 150, 151, 154 melanogaster 14, 59, 61, 69, 74–5, 128, 137, 138, 150, 158, 161, 162, 172, 173, 179, 197 serrata 59, 61 similans 150 subobscura 95, 161–2, 225

Eleutherodactylus 64 Embryonopsis halticella 60 Erigeron candensis 129 Escherichia coli 14, 71–2, 74, 95, 128 Eucalyptus pauciﬂorac 147

Fibrocaspa japonica 67 Flaveria bidentis 129 Fundulus heteroclitus 42, 132, 150

Gadus morhua 57 Gambusia holbrooki 132, 172 Gasterosteus aculeatus 132 Gerrhonotus multicarinatus 41 Girella laevifrons 104 Gryllus pennsylvanicus 154 Gymnogongrus antarcticus 129

Haliotis 137 Halozetes 150 Helianthis mollis 129 Heliotropium curassavicum 147 Hemidactylus frenatus 107 Heterandria formosa 85 Hoplodactylus maculates 107 Hyla 64 Hyperaspis notata 56, 57 Hyphantria cunea 225

Idotea baltica 168 Inachis io 90

Kallymenia antarctica 129 Koliella anatarctica 39

Lampropeltis triangulum 110 Larrea divaricata 131 Ledum groenlandicum 147, 184, 186 Lepidodactylus lugubris 107 Lepomis macrochirus 113 Limnodynastes peronii 62, 133, 134 Locusta migratoria 105 Lycaena hippothoe 170 tityrus 60 Lymnaea stagnalis 106 283

284

SPECIES INDEX

Mabuya multifasciata 107 Macoma balthica 57 Macrobiotus harmsworthi 132 Macrobrachium acanthurus 186–7 Macrosiphum euphorbiae 35 Malpolon monspessulanus 107 Menidia menidia 67, 68, 77 peninsulae 67, 68, 78 Micropogonias undulatus 57 Micropterus dolomieui 133 Mnium arizonicum 71 spinulosum 71 Morelia spilota 104 Myoxocephalus scorpius 133 Mytilus edulis 126 Myzus persicae 35

Natrix natrix 107 Nerium oleander 131 Nerodia rhombifer 60, 107 sipedon 103, 104, 107 Notothenia neglecta 44

Oncorhynchus kisutch 133 Orchestia gammarellus 60, 150 Ostrea edulis 132 Oxyria digyna 147

Pagothenia borchgrevinki 133 Panicum coloratum 129 Panstrongylus megistus 104 Pantherophis guttatus 104 obsoleta 103, 104 Papilio canadensis 57 Pararge aegeria 39 Penicillium bilaii 225 Petrolisthes 58, 60, 150, 152 agassizir 61 armatus 61 cabrillor 61 cinctipes 61, 152 crenulatus 61 eriomerus 61, 152 galathinus 61 gracilis 61, 152 granulosus 61

haigae 61 hirtipes 61, 152 holotrichus 61 laevigatus 61 manimaculis 61 platymerus 61 sanfelipensis 61 tridentatus 61 tuberculatus 61 tuberculosus 61 violaceus 61 Phyllophora ahnfeltioides 130 Pieris rapae 78, 79, 81–3, 159 Pisum sativum 130 Plantago euryphylla 148 lanceolata 148 major 148 Podacarus auberti 150 Podacris melisellensis 112 muralis 107 sicula 112 Polygonia c-album 90 Porcellio laevis 60 Potamogeton pectinatus 68 Pseudacris triseriata 60 Pseudemoia entrecasteauxii 112 Puntius schwanenfeldii 133

Rana 66 pipiens 134 temporaria 56, 57, 161 Ruppia drepanensis 130 Rutilus rutilus 104

Salmo salar 42, 104 trutta 202 Salvelinus namaycush 104 Scaptodrosophila 60 Scathophaga stercoraria 171, 172 Sceloporus 151 aeneus 107 arenicolus 228 bicanthalis 107 gadoviae 107 graciosus 96, 221 grammicus 107 jarrovi 106, 107

merriami 109 occidentalis 96 scalaris 107 undulatus 41, 107, 112, 161, 177, 193, 221, 223 virgatus 107 Scyliorhinus canicula 116 Simocephalus vetulus 178, 226–7 Sorghastrum nutans 130 Spinacia oleracea 130 Stellio stellio 62

Takydromus septentrionalis 39 sexlineatus 39, 40 wolteri 39, 40 Taurulus bubalis 133 Tegula 151 Thamnophis elegans 107 sirtalis 104, 107 Tidestromia oblongifolia 131 Tiliqua rugosa 116 Torulomyces lagena 225 Trebouxia irregularis 67 Triatoma infestans 104 Trichoderma koningii 225 Trichogramma carverae 137 Trimeresurus stejnegeri 104 Triticum aestivum 130 Triturus dobrogicus 104, 134

Uma notata 41 Umbilicus rupestris 225 Uromastyx philbyi 107 Urosaurus ornatus 107

Valonia utricularis 59, 130, 148 Vanessa atalanta 90 Vicia dumetorum 67 pisiformis 67

Xantusia vigilis 134, 136 Xenopus laevis 134 Xylia xylocarpa 130

Zootoca vivipara 60, 102, 107, 116–17

Subject Index

acclimation developmental 126–31, 141–3, 146–8 reversible 126–7, 131–5, 143–5, 149 see also thermal acclimation acquisition tradeoffs 6–7, 137, 195 see also tradeoffs additive genetic variance 8–11, 149–51, 172–3, 229–32 age at maturity experimental evolution of 161–2 optimal reaction norms 162–70 thermal clines 160–1 thermal plasticity 158–60 Akaike’s information criterion (AIC) 40 allelic models 10–13, 209 allocation tradeoffs 6–7, 137, 157–8, 179–80, 195 see also tradeoffs altitudinal clines 27–9 thermal performance curves 54–6 anabolism 167–8 analysis of variance (ANOVA) 11 artiﬁcial neural networks 33 operative temperature model 33 artiﬁcial selection 11–12, 14 experimental studies of thermal adaptation 71–4 behavioral cascade 205 behavioral thermoregulation 96 beneﬁcial acclimation hypothesis 127–35 developmental acclimation 127–31 reversible acclimation 131–5 beneﬁts of acclimation 127–35 of thermoregulation 91–6, 99, 100–2, 105–7 Bergmann’s clines 159–61 see also temperature–size rule body size temperature–size rule 158–61, 162, 164–70

thermal constraints 166–9 see also offspring size; size at maturity body temperature core and surface temperature 20–1, 96–7 preferred 89–90, 92 evolution of by indirect selection 197 mismatch with thermal optima 189–93 steady-state temperature 20–2 see also temperature; thermoregulation breeding experiments 11 catabolism 167–8 cell membranes see membrane structure chill-coma recovery 58–61 climate envelope model 217 global change 214–15 local change 216 microclimate 21, 100, 108, 112–13 regional change 215 space diagram 22 see also global warming clines see thermal clines coadaptation 181–2, 196–8 constraints on 196 thermal physiology and thermoregulatory behavior 186–95 predicting coadapted phenotypes 193–5 thermal sensitivity and thermal acclimation 182–6 see also thermal adaptation coarse-grained environments 140 coevolution during global warming 233 thermal breadths 210–11 thermal optima 207–10 gene ﬂow and 211–12

cogradient variation 67, 68 cold tolerance 59–60 see also hardening; thermal acclimation common garden experiments 57, 59, 63, 66, 92, 161 comparative analysis 15–18 competition during thermoregulation 201–3 thermoregulatory costs 111–12 correlative models 219–20 costs of generalization see specialist–generalist tradeoffs of phenotypic plasticity 7 of thermal acclimation 135–40 of thermoregulation 96–8, 99–102, 107–18 countergradient variation 67 critical thermal limits 41, 44, 52, 134, 186–7 critical thermal maximum 36, 38, 41, 44–5, 57–8, 187–8 critical thermal minimum 36, 41, 44, 47, 187

desiccation risk, thermoregulatory costs 116 development, thermal sensitivity of, 46, 66 developmental acclimation 126–31 comparative studies 146–8 optimal 141–3 developmental rate 46, 66 distribution geographic 219–23 ideal free 201–7 drift see genetic drift dynamic optimization 7

egg size see offspring size endothermy 119–25 evolutionary origins 119–22 optimal thermoregulation 122–5 285

286

SUBJECT INDEX

energetic costs 136–9 energy allocation 167 to reproduction 171–9 to growth 164 energy gain 103–4 optimal offspring size and 176–7 environmental symmetry 212 environmental temperature see temperature enzymes limitation on performance 45–6 thermal effects on 41–3 evaporative cooling 20–2, 97 evolutionarily stable strategy 200 evolutionary games 199 life-history evolution 212–13 optimal performance curves 207–12 optimal thermoregulation 200–7 evolutionary thermal biology 1–3 empirical tools 13–18 laboratory selection experiments 71–5 reciprocal transplant experiments 70–1, 74 theoretical approaches 6–13 experimental evolution 14, 16 age and size at maturity 161–2 offspring size 172–4 extrinsic mortality 195–6

fecundity genetic model 83–4 optimal performance curves 51–3 feedback regulation 139–40 feedforward regulation 139–40 feeding energy gain 103–4 missed opportunities, thermoregulatory costs 116–18 fever 105–6 ﬁne-grained environments 140 ﬁtness 7–8 interacting traits 181–2 performance and 48, 76–9, 157 ﬁtness curves 35–6, 47 ﬁtness landscape 6, 14 ﬁxed traits 3–4 food quality 103–4 frequency-dependent optimization 7 frequency-dependent selection 200 functional constraints see tradeoffs

game theory 199 evolutionarily stable strategy 200

life-history evolution 212–13 optimal performance curves 207–12 optimal thermoregulation 200–7 Gaussian function 49, 141, 207–8, 210–11 gene ﬂow coevolution of thermal optima and 211–12 constraints on the evolution of acclimation 152–4 constraints on thermal adaptation 85–7 global warming and 233–6 genetic algorithm see allelic models genetic constraint 196 genetic correlation 41, 75, 83, 121, 152, 186, 196 genetic drift 8, 13, 16, 85, 229–31 genetic engineering 14 genetic variance 8–10 acclimation responses 151 additive 8–11, 149–51, 172–3, 229–32 constraints on the evolution of acclimation 149–52 constraints on thermal adaptation 79–85 variance–covariance function 81 geographic range shifts 217 global warming 214–15 adaptation to directional thermal change 224–32 performance breadth 231–2 thermal optimum 229–31 thermoregulation 227–9 coevolution and 233 evolutionary consequences of gene ﬂow 233–6 observed responses 216–18 ecological interactions 217–18 geographic range 217 phenology 216–17 primary productivity 218 prediction of responses 218–24 correlative versus mechanistic models 219–20 differential responses of populations and species 223–4 mechanistic models 220–3 thermal games and 232–3 see also climate growth rate 66 thermal sensitivity of 66–8

hardening 126, 137–9 comparative studies 149 genetic variance 151 heat exchange 19, 20–1 heat ﬂux 21, 31–2 conduction 20–1 convection 20–3 evaporative cooling 20–2, 97 radiation 20–3 heat hardening see hardening heat tolerance 11–12, 59–60 see also hardening; thermal acclimation heat-shock proteins 137–40 heritability 8, 196 hibernation 122, 123, 125 hotter is better 94–5, 192–3, 194 hypothermia 122 hypoxia 104–5

ideal free distribution 201–7 independent contrasts 17, 18 individual-based model 200, 233 infection, thermoregulatory beneﬁts 105–6 information theory 40 intrinsic mortality 195 irreversible phenotypes 140

juvenile mortality 164–6

knockdown temperature 57 knockdown time 57

labile traits 3–4 lactate dehydrogenase (LDH) 42 latitudinal clines 23–7 thermal performance curves 54–6 lethal temperature 58, 60 life history age and size at maturity 158–70 experimental evolution of 161–2 optimal reaction norms 162–70 thermal clines 159–61 thermal plasticity 158–60 coadaptation 195–6 evolution in a thermal game 212–13 performance and 157–8 thermal adaptation 157–8 locomotor performance jumping distance 62–3 sprinting speed 62–3, 65, 189–91

SUBJECT INDEX

locomotor performance (Cont.) swimming speed 42, 62–4, 78 thermal sensitivity of 62–6

Malthusian parameter (r) 48, 157, 168 see also ﬁtness mathematical models 5–6 local variation in operative temperature 29–30 mechanistic models 219–20 responses to global warming 220–3 membrane structure 43–4 ﬂuidity 43–4, 136 metabolic heat production 119–21 see also endothermy metabolic theory of ecology 2, 94, 192 microclimate 21, 100, 108, 112–13 microhabitat 211, 232 migration 85–7, 152–3 vertical migration 120, 206–7 see also gene ﬂow migrational meltdown 85 models 2–3, 5–6, 13 allelic 10–13, 209 correlative 219–20 evaluation 6, 13 graphical 5 mathematical 5–6 mechanistic 219–23 optimality 6–8, 13 quantitative genetic 8–10, 13, 79–85, 149 verbal 5 mortality extrinsic mortality 195–6 intrinsic mortality 195 juvenile 164–6 multivariate selection theory 8–9, 81 mutation 7, 8, 13, 42, 81–5

Nash equilibrium 200 natural selection 7–8 ﬁtness landscape and 6 multivariate selection theory 8–9, 81 quantiﬁcation 14, 15 see also selection gradients net photosynthetic rate 67–8, 185 net reproductive rate 48, 163 see also ﬁtness null hypothesis 6

offspring number 174–8 offspring size 171–9

experimental evolution of 172–4 optimal 174–8 direct thermal effect 174–8 indirect thermal effects 176–8 optimal variation 178–9 thermal clines 171–2 thermal plasticity 171 see also body size operative environmental temperature 20–2, 33–4 frequency distribution 109–11 global patterns 23–9 quantifying local variation 29–33 see also temperature optimal performance curves 47–54 fecundity and related performances 51–3 survivorship and related performances 47–51 see also performance curves optimality models 6–8, 13, 162, 179 approaches 163–4 genetic constraints 79–85 phenotypic constraints 75–6 thermoregulation 98–102 types of 7 oxygen limitation 44

parasitism, thermoregulatory costs 112–16 performance 36 ﬁtness and 48, 76–9, 157 life history and 157–8 measures of 37 thermal limits 38, 56–61 performance curves 36–40 evolutionary games and 207–12 laboratory selection experiments 71–5 modeling genetics 209 natural patterns 54–70 development 66 differences from predicted patterns 69–70 growth 66–8 locomotion 62–6 reproduction 68–9 survivorship 56–62 optimal acclimation of 140–6 evidence of 146–9 optimal developmental acclimation 141–3 optimal reversible acclimation 143–5 optimal performance curves 47–54

287

fecundity and related performances 51–3 survivorship and related performances 47–51 proximate mechanisms 41–7 membrane structure 43–4 oxygen limitation 44 enzymatic function 41–3, 45–6 reciprocal transplant experiments 70–1, 74 survivorship during development 46 template model of variation 74 thermoregulatory precision and 102–5 tradeoffs 47, 52–3, 75–6, 77, 78 see also thermal adaptation; thermal sensitivity phenology, climate change and 216–17 phenotypic constraints 75–6 phenotypic engineering 14 phenotypic plasticity 3–4, 149 costs of 7 see also thermal plasticity photosynthesis net photosynthetic rate 67–8, 185 thermal acclimation 128, 129–30, 131, 147–9 thermal sensitivity of 67–8 phylogenetic comparative analysis see comparative analysis physical models, local variation in operative temperature 31–2 plasticity see phenotypic plasticity population growth, thermal constraints on 169–70 predation during thermoregulation 203–5, 206–7 thermoregulatory costs 112–16 preferred body temperature 89–90, 92 evolution of by indirect selection 197 mismatch with thermal optima 189–93 pregnancy, thermoregulatory beneﬁts 106–7 prey ideal free distribution 203–7 predation during thermoregulation 203–5 primary productivity, climate change and 218 proteins, thermal effects on 41–3

288

SUBJECT INDEX

quantitative genetic models 8–10, 13, 79–85, 149 based on multivariate selection theory 81 estimating parameters of 11–12 fecundity and related performances 83–4 predictions of 84–5 survivorship and related performances 81–3 range limit 221 reaction norms see thermal reaction norms reciprocal transplant experiments 70–1, 74 recovery time 57 refuge 112–16, 205 reproduction missed opportunities, thermoregulatory costs 116–18 thermal sensitivity of 68–9 reproductive allocation 171–9 optimal offspring size and number 174–8 reproductive rate 48, 163 see also ﬁtness reversible acclimation 126–7, 131–5 comparative studies of 149 optimal 143–5 reversible phenotypes 140 selection gradients 9–10, 81 measurement of 14, 15 see also natural selection set point limit 93, 102, 117 shuttle box 108 size at maturity experimental evolution of 161–2 optimal reaction norms 162–70 thermal clines 159–6 Bergmann’s clines 159–61 thermal plasticity 158–60 see also body size solar radiation 23, 24 specialist–generalist tradeoffs 7, 47, 52–3, 75–6, 77, 137 adaptation to directional change and 231–2 see also tradeoffs static optimization 7 statistical models, local variation in operative temperature 32–3 strong inference 6, 177 survivorship genetic model 81–3

modeling thermal sensitivity 54–62 optimal performance curves 47–51

temperature 1, 19, 35 competition for 111–12 effects on enzymes 41–3 global mapping of 25–6 strategies for coping with heterogeneity 1–2 thermal reaction norms 3–4, 35–6, 38–40 see also body temperature; operative environmental temperature; thermal clines; thermal heterogeneity; thermal plasticity temperature–size rule 158–61, 162, 164–70, 179 theoretical biology 5–6 evolutionary approaches 6–13 theory 5–6 complementarity of 13 see also theoretical biology thermal acclimation 126, 155–6 beneﬁcial acclimation hypothesis 127–35 coadaptation with thermal sensitivity 182–6 costs of 135–40 energetic demands 136–9 interaction between 139–40 time lags 139 developmental 126–31, 141–3, 146–8 ecological relevance 154–5 environmental stability and 146–9 evolutionary constraints 149–54 gene ﬂow 152–4 genetic variance and covariance 149–52 genetic variance 151 optimal acclimation of performance curves 140–6 evidence of 146–9 patterns of 126–7 reversible 126–7, 131–5, 143–5, 149 thermal adaptation gene ﬂow 85–7 genetic variation 79–85 laboratory selection experiments 71–5 life history 157–8 reciprocal transplant experiments 70–1, 74 to directional thermal change 224–32

adaptation of performance breadth 231–2 adaptation of thermal optimum 229–31 adaptation of thermoregulation 227–9 see also coadaptation; evolutionary thermal biology; thermal sensitivity thermal clines 70–5 age and size at maturity 159–61 Bergmann’s clines 159–61 altitudinal clines 27–9 laboratory selection experiments 71–5 latitudinal 23–7 offspring size 171–2 reciprocal transplant experiments 70–1 thermal performance curves 54–6 see also temperature; thermal heterogeneity thermal coadaptation see coadaptation thermal equilibrium 19 thermal games 199 life-history evolution 212–13 optimal performance curves 207–12 optimal thermoregulation 200–7 thermal gradients 92, 103 thermal heterogeneity 19, 34, 70, 145, 154, 157 heterogeneous warming 233–6 quantifying local variation 29–33 see also temperature; thermal clines thermal limits 38, 56–61 acclimation of 134–5 see also critical thermal limits thermal mapping 25, 26, 32 thermal optimum 94–6 acclimation of 146 adaptation to global warming 229–31 coevolution between species 207–10 gene ﬂow and 211–12 mismatch with preferred temperature 189–93 thermal performance curves see performance curves thermal plasticity age and size at maturity 158–60 offspring size 171 see also phenotypic plasticity thermal preferences 89–90, 92 measurement of 92

SUBJECT INDEX

thermal reaction norms 3–4, 35–6, 38–40 age and size at maturity 162–70 evolution of 9–10 optimal 7 thermal sensitivity 1 coadaptation with thermal acclimation 182–6 of development 46 genetic variance of 84–5 patterns of 35–40, 54–70 development 66 growth 66–8 locomotion 62–6 reproduction 68–9 survivorship 56–62 proximate mechanisms 41–7 specialist–generalist tradeoffs 47, 52–3 strategies of 1–2 see also performance curves; thermal adaptation thermal tolerance, indices of 56–8 see also thermal sensitivity thermodynamic constraints see hotter is better thermoregulation 1, 70, 88, 98 adaptation to global warming 227–9 after feeding 103 behavioral 96

coadaptation 186–96 beneﬁts of 91–6, 99, 100–2, 105–7 during infection 105–6 during pregnancy 106–7 precision and 102–5 costs of 96–8, 99–102, 107–18 competitive interactions 111–12 estimating energetic costs 112–13 desiccation risk 116 interactions between different costs 118 missed opportunities for feeding/reproduction 116–18 precision and 107–11 predation/parasitism 112–16 endothermic 119–25 evolutionary origins 119–22 optimal thermoregulation 122–5 evolutionary games and 200–7 competition during thermoregulation 201–3 predation during thermoregulation 203–5, 206–7 indices of 93 accuracy of thermoregulation 90, 93, 110 effectiveness of thermoregulation 90, 93, 103 thermal quality of the environment 93, 109 optimal 98–102, 200–7

289

quantifying patterns of 88–91 strategies 1–2 thermosensitivity see thermal sensitivity costs of time lags 139 tolerance curves 35, 47 see also ﬁtness curves torpor 122, 123–4 tradeoffs 6–7, 77, 135–6, 137, 195 acquisition 6–7, 137, 195 allocation 6–7, 137, 157–8, 179–80, 195 case study 78 specialist–generalist 7, 47, 52–3, 75–6, 77, 137 adaptation to directional change and 231–2 transplant experiments see reciprocal transplant experiments trypsin 42

vertical migration 120, 206–7

warmer is better 94 see also hotter is better wet operative temperature 22

(a)

(b)

(c)

(d)

58 46 34 21 9

Plate 1 Operative temperatures available to small lizards in the Mescalaro sand dunes (New Mexico, USA) change dramatically throughout the morning on a summer’s day (M. W. Sears, G. S. Bakken, M. J. Angilletta, and L. A. Fitzgerald, unpublished observations). (a–d) Each map depicts an operative thermal environment over an area of 460 × 320 m at a resolution of 0.5 m. (a) 7 a.m.; (b) 9 a.m.; (c) 11 a.m.; (d) 1 p.m. See Page 34.

(a)

(b)

(c)

Frequency

60 50 40 30 30 40 50 60 Operative temperative (°C)

Plate 2 (a) A bimodal distribution of operative temperatures was used to simulate the energetic cost of behavioral thermoregulation. (b and c) Depending on the simulation, these temperatures were distributed in one large clump or in many small patches. See Page 112.

–1 to –0.7

–0.7 to –0.3

> –0.3

< –0.3

0.3 to 0.7

0.7 to 1

Plate 3 Environmental temperatures have changed at different rates throughout the world. This map shows rates of change (◦ C per decade) between 1976 and 2000, relative to normal temperatures for 1961–1990 (see Box 2.1). Reprinted by permission from Macmillan Publishers Ltd: Nature (Walther et al. 2002), © 2002. See Page 215.